Optimal Categorisation: emergent gender in Oceania

The possessive classifier systems of some Oceanic languages show signs of shifting towards fixed gender systems. In typical classifier systems, a noun can occur with different classifiers, depending on how the possessed item is used by the possessor (Lichtenberk, 1983). For example, nompui ‘pig’ in Lewo (Vanuatu) occurs with either the food or the general classifier.

Lewo (typical system)

In marked contrast, North Ambrym’s (Vanuatu) cognate for pig – bu – occurs only with the food classifier (2a), not the general classifier (2b):

North Ambrym (innovative system)

We argue that North Ambrym’s innovative system shows some properties of a gender system: a noun occurs with a particular classifier regardless of contextual interactions. We ask whether gender systems can indeed emerge from possessive classifiers in this way. If so, we must then uncover how and why languages would relinquish a useful, meaningful classificatory system, and adopt a rigid, apparently less motivated gender system.

We designed a suite of experiments to compare possessive classifier systems in six representative Oceanic languages: Merei, Lewo, Vatlongos, North Ambrym (Vanuatu), Nêlêmwa and Iaai (New Caledonia). Each has a different inventory of classifiers, from a simple two-way distinction (Merei) up to a more complex inventory of 23 (Iaai). The classifier systems in these languages also represent varying degrees of informativeness: from transparent semantic motivation to opaque assignment. Effective categorisation needs to be informative to maximise communicative efficiency, but also needs to be simple, to minimise cognitive load (Hawkins, 2004). Our sample languages allow us to investigate the trade-off between these two principles of informativeness and simplicity.

We discuss results from three experiments, involving data from 122 speakers across our sample languages:

1.  Free-listing, where participants are asked to list the nouns appropriate for use with each classifier. This establishes central members of a classifier’s semantic domains.
2.  Video vignettes, which depict typical and atypical interactions with different items. These determine whether speakers are free to use different classifiers (as generally believed) or whether assignment of nouns to classifiers is rigid.
3. Card sorting, where participants are asked to sort pictures depicting entities into groups, depending on which classifier the corresponding noun can occur with. This task reveals how speakers categorise relevant nouns and whether conceptual groupings map onto the classifier inventory available to the speakers.

Data from the three experiments have been analysed in different ways. The results from the free-listing and video vignette experiments show that the sample languages differ in how rigidly the nouns are assigned to particular classifiers, with North Ambrym having the most rigid assignment. When we combine the results from the free-listing and the card sorting experiments, we see a typology of classifier systems: classifiers that have large, open membership dominate the smaller, less open classifiers.

The results suggest that the classifier systems tested represent different stages of grammaticalization, from noun to classifier to gender marker. They reveal intriguing inter-speaker variation in classifier choice, which helps establish the relative optimality of each system.

References:
Hawkins, John. A. 2004. Efficiency and Complexity in Grammars. Oxford: Oxford University Press.
Lichtenberk, Frantisek. 1983. Relational Classifiers. Lingua 60(2-3):147–176.