How to measure the optimality of word or gesture order with respect to the principle of swap distance minimization

Abstract: The structure of all the permutations of a sequence can be represented as a permutohedron, a graph where vertices are permutations and two vertices are linked if a swap of adjacent elements in the permutation of one of the vertices produces the permutation of the other vertex. It has been hypothesized that word orders in languages minimize the swap distance in the permutohedron: given a source order, word orders that are closer in the permutohedron should be less costly and thus more likely. Here we explain how to measure the degree of optimality of word order variation with respect to swap distance minimization. We illustrate the power of our novel mathematical framework by showing that crosslinguistic gestures are at least $77\%$ optimal. It is unlikely that the multiple times where crosslinguistic gestures hit optimality are due to chance. We establish the theoretical foundations for research on the optimality of word or gesture order with respect to swap distance minimization in communication systems. Finally, we introduce the quadratic assignment problem (QAP) into language research as an umbrella for multiple optimization problems and, accordingly, postulate a general principle of optimal assignment that unifies various linguistic principles including swap distance minimization.

• The paper introduces a framework to quantify optimality via swap distance minimization across permutation spaces.
• It employs normalization techniques and the Quadratic Assignment Problem to derive a robust optimality score for word and gesture arrangements.
• Empirical analysis on gestural data shows consistent evidence of contiguous, adjacent, and optimal orderings across diverse languages.

Measuring the Optimality of Word/Gesture Order via Swap Distance Minimization

Introduction

The study introduces a formal mathematical framework for quantifying the optimality of word or gesture order based on swap distance minimization within the permutohedron graph. The permutohedron, in this context, models the space of all possible permutations of a sequence (such as SOV, SVO, etc.) where edges represent minimal edit operations—specifically, adjacent swaps. The key hypothesis is that, in human language and gesture, observed permutations tend to minimize swap distance relative to a source or dominant order, operationalizing "cost" as the swap distance within this discrete structure.

The paper provides a rigorous normalization procedure for measuring this optimality, investigates theoretical consequences of swap distance minimization for the distributional structure of word orders, and demonstrates the application of the framework using empirical data from crosslinguistic gestural experiments.

Mathematical Framework and Normalization

The core metric, average swap distance $(d)$, is a distribution-weighted mean of swap distances across all pairs of permutations, generalizing simple entropy-like diversity measures by integrating topological structure. Minimizing $(d)$—so that frequently observed orders cluster closely on the permutohedron—constitutes the basic optimality criterion.

Prior work established that $(d)$ in natural languages is generally lower than expected by chance, but lacked a robust normalization capturing "distance to optimality" in a way stable under null empirical distributions. This paper addresses this by adapting similarity score normalization techniques from clustering and language optimization research, most notably the Hubert-Arabie adjustment. The final "degree of optimality" score is:

$\eta_2 = \frac{(d)_r - (d)}{(d)_r - (d)_{\min}}$

Here, $(d)_r$ is the expected value under random permutation of order probabilities (the permutation null), and $(d)_{\min}$ is the minimum attainable value over all possible assignments of probabilities. This normalization ensures that $\eta_2 = 1$ in the optimal case and $\eta_2 = 0$ under the null hypothesis, providing consistency for typological comparisons in both rigid and flexible word-order systems.

Computing $(d)_{\min}$ is shown to be a special case of the well-known Quadratic Assignment Problem (QAP), with the additional structure that the cost matrix has rank 1 (it is the outer product of the probability vector with itself), and the distance matrix derives from the permutohedron. The framework is general enough to subsume other optimization principles widely discussed in computational linguistics, such as dependency distance minimization and compression, under an overarching "optimal assignment principle."

Empirical Application to Crosslinguistic Gestures

Using experimental data from Futrell et al. (2015), the study computes order frequency distributions for SOV permutations produced in gestural communication by speakers of English, Russian, Irish, and Tagalog, under reversible and non-reversible event conditions. For each language-condition pair, they report:

• Number of non-zero probability orders ($m$)
• Dominance index $(d)$0 (Simpson's index)
• Observed $(d)$1, random expectation $(d)$2, and $(d)$3
• The degree of optimality $(d)$4

Results demonstrate that in all 8 language-condition cases, $(d)$5, and in 4 cases, $(d)$6, indicating full optimality. Wilcoxon signed-rank tests confirm that the observed $(d)$7 values are significantly lower than random, and Poisson binomial tests show that the number of fully optimal arrangements is higher than expected by chance. Notably, the study documents that arrangements with full optimality also exhibit key epiphenomena—such as "radiation from the most likely order," adjacency of the two highest-probability orders, and contiguity (all non-zero probability orders form a single path in the permutohedron). The pervasiveness of contiguous sets is statistically unlikely to result from chance.

Theoretical Implications

Structure and Typological Consequences

The framework entails several strong structural predictions for word/gesture order distributions under swap distance minimization:

• Radiation: Probability monotonically decreases with increasing swap distance from the dominant order.
• Adjacency: The two most frequent orders must be directly connected in the permutohedron.
• Contiguity: Non-zero probability orders always form a contiguous path.

These are not presumed as independent grammatical universals, but rather as inevitable mathematical consequences ("epiphenomena") of swap distance minimization. The analysis generalizes to both rigid and flexible word-order languages and applies across modalities, as shown by consistent results in both gestural and spoken communication.

Additionally, by formulating the problem as a QAP, the study subsumes swap distance minimization, dependency length minimization, and compression (in the sense of code length or linguistic economy) within a unified optimal assignment principle—a broader theoretical umbrella that also encompasses optimization problems in economics and biology.

Methodological Innovations

• The normalization of $(d)$8 with respect to the permutation null enables typological comparisons across languages with different degrees of word order flexibility.
• The use of the QAP and explicit algorithms for $(d)$9 calculation introduces a new level of methodological rigor to permutation-based typological analysis.

Implications for Linguistic Theory

The approach challenges prevailing typological methodologies that focus only on the mode or "dominant" order (type-based typology), highlighting the necessity of considering the full probability distribution (token-based typology) and its structure with respect to the permutohedron, which captures the inter-order relationships.

The findings suggest that typological universals or tendencies (e.g., SOV/SVO dominance, adjacency of dominant orders, contiguity) can be derived from swap distance minimization without requiring additional stipulations. Deviant or less common word order patterns in specific languages are not anomalous in this framework, but represent alternate instantiations that respect the same underlying constraint.

Future Directions

The research framework generalizes naturally to permutation sets with $(d)$0 elements, though computational costs for computing $(d)$1 increase rapidly. The interplay between swap distance minimization and other functional motivations (e.g., information structure, cognitive constraints, event semantics) warrants further empirical and mathematical exploration.

A proposed line of work is to examine the universality of the optimal assignment principle in a broader set of linguistic and communicative domains, including spontaneous gesture genesis, child language acquisition, and comparative animal communication.

Conclusion

This study provides a mathematically principled, empirically validated framework for measuring the optimality of word or gesture order under the principle of swap distance minimization in the permutohedron. With robust normalization, demonstration of application to gestural data, and integration with the QAP, the work significantly enhances the methodological toolkit for computational typology, and for the modeling of the interface between cognitive pressures and typological distributions. Theoretical ramifications extend across linguistics and cognition, offering a unifying account for several longstanding optimization principles in language and communication.

---

Reference: "How to measure the optimality of word or gesture order with respect to the principle of swap distance minimization" (2604.01938)