Systems of nominal classification act as a functional means of categorisation, yet the number and type of categorieswithin these systems vary considerably across languages. The impact of vastly different classification systems on thecognitive representations of concepts is intriguing. We designed a suite of experiments to compare classifier systems in sixOceanic languages, chosen because their inventory of classifiers ranges from two to 23. Effective categorisation needs tobe informative to maximise communicative efficiency, but also simple to minimise cognitive load. Our sample languagesallow us to investigate the trade-off between the two principles of informativeness and simplicity to shed light on therelative optimality of their classification systems. Results from 122 participants across three experiments (free listing,card sorting, video vignettes) indicate that cognitive salience varies as a function of classifier inventory. We discuss theimplications of these results for the nature of nominal classification.

Effective categorisation should be simple, to minimise cognitive load, and informative, to maximise communicative effi-ciency. Nominal classification systems (gender, classifiers) are a functional means of categorisation that vary enormouslyacross languages, revealing a trade-off between simplicity and informativeness. Closely related Oceanic languages ofMelanesia show staggering variation in their number and type of classifiers. How does the Iaai language carve up nounsinto 23 semantic groups whilst the Merei language uses only two; and what implications do these vastly different systemshave for the cognitive representations of their related concepts? We combined typological enquiry and psycholinguisticexperimentation (free listing, card sorting, video vignettes, possessive labelling, eye tracking, storyboards, category train-ing) comparing nominal classification systems in six Oceanic languages of Vanuatu and New Caledonia. We discuss howthese experiments uncover the nature of nominal classification systems, comparing objective data across languages andexperimental contexts to reveal a model for optimal categorisation.