The communicative role of adjectives: evidence from Hungarian

Abstract

Language is shaped by pressures of communicative efficiency. In a paper published in 2018, Dye and colleagues proposed that adjectives serve a communicative function by smoothing information density: formulaic adjectives make low-frequency nouns easier to predict, reducing processing difficulty. I test this communicative account against a sampling account, which claims that apparent smoothing emerges as a sampling artefact of Zipfian frequency distributions rather than from functional pressures. Using a nine-billion-word corpus of Hungarian, a language without grammatical gender and with information-structure-driven word order, I replicate Dye et al.’s findings: low-frequency nouns show higher modification rates, lower adjective entropy, and stronger collocational effects. However, numerals, which arguably lack the semantic flexibility to serve as formulaic supports, pattern nearly identically to adjectives across most measures. A permutation test reveals that the communicative signal is real but small: it is found in the shape of the adjective distribution rather than its support size, with low-frequency nouns showing greater concentration of adjective use relative to their sampling baseline. The sampling artefact establishes the baseline distributional patterns; communicative pressures shape the residual.