Hierarchies in Dictionary Definition Space

A dictionary defines words in terms of other words. Definitions can tell you the meanings of words you don’t know, but only if you know the meanings of the defining words. How many words do you need to know (and which ones) in order to be able to learn all the rest from definitions? We reduced dictionaries to their “grounding kernels” (GKs), about 10% of the dictionary, from which all the other words could be defined. The GK words turned out to have psycholinguistic correlates: they were learned at an earlier age and more concrete than the rest of the dictionary. But one can compress still more: the GK turns out to have internal structure, with a strongly connected “kernel core” (KC) and a surrounding layer, from which a hierarchy of definitional distances can be derived, all the way out to the periphery of the full dictionary. These definitional distances, too, are correlated with psycholinguistic variables (age of acquisition, concreteness, imageability, oral and written frequency) and hence perhaps with the “mental lexicon” in each of our heads.

💡 Research Summary

The paper investigates the structural organization of a dictionary by treating definitions as a directed network, where each word is a node and an edge points from a defined word to each word used in its definition. By constructing this “definition graph,” the authors ask a fundamental question: how many words must a learner already know in order to be able to acquire the meanings of all other words solely through definitions?

To answer this, they identify a minimal set of nodes that, once known, allow every other node to be reached via directed paths. This set is called the grounding kernel (GK). Empirically, across several large English dictionaries, the GK comprises roughly ten percent of the total lexicon. In other words, knowing just this small subset is sufficient to define the remaining ninety percent of words.

A deeper graph‑theoretic analysis reveals that the GK is not a homogeneous block but contains an internal hierarchy. The largest strongly connected component within the GK is designated the kernel core (KC). Words in the KC mutually define each other, forming cycles, whereas the rest of the GK lies in a surrounding layer that points outward toward the periphery of the full dictionary. The authors introduce a metric called definitional distance, measured as the shortest number of directed edges from any word to the KC. This yields a layered map extending from the KC through the surrounding GK layer to the outermost words that are farthest from the core.

To connect these structural findings with human language processing, the study correlates definitional distance and kernel membership with a suite of psycholinguistic variables: age of acquisition (AoA), concreteness, imageability, oral frequency, and written frequency. Using established norms (e.g., Kuperman et al. for imageability, Brysbaert et al. for frequency), they show that KC words are learned earliest, are the most concrete and imageable, and occur most frequently in both spoken and written language. As definitional distance increases, these properties systematically decline, indicating that the periphery consists of later‑acquired, more abstract, and less frequent vocabulary. This pattern provides empirical support for the “grounding hypothesis,” which posits that human mental lexicons are built upon a core of concrete, high‑frequency words that serve as semantic anchors for the rest of the vocabulary.

Beyond theoretical insight, the authors discuss practical implications. In language education, targeting the GK—especially the KC—early in instruction could dramatically accelerate overall vocabulary acquisition, because learners would then have the necessary scaffolding to infer meanings of many other words. Adaptive learning software could exploit the definitional distance metric to present new words in an order that respects the learner’s current knowledge state, thereby minimizing cognitive load. In natural language processing, incorporating the GK as a seed lexicon may enable more efficient construction of semantic networks, especially in low‑resource settings where full lexical coverage is unavailable. Finally, the hierarchical model offers a new lens for studying dialectal variation, technical jargon, and language change, as shifts in the GK or KC could reflect deeper changes in how a language is grounded in experience.

In summary, the paper demonstrates that dictionaries possess a robust hierarchical architecture: a small, highly interconnected kernel core, surrounded by a layer of words that directly depend on the core, and a peripheral fringe that is increasingly abstract and less frequent. This architecture mirrors psycholinguistic patterns of word learning and usage, suggesting that the mental lexicon is organized in a comparable fashion. The findings bridge graph theory, psycholinguistics, and applied language technology, opening avenues for more principled vocabulary teaching methods and more efficient computational models of meaning.