On the Relationship between Polysemy and Type Frequency

Abstract

Zipf and others in his wake examined the relationship between polysemy and token frequency but largely neglected the relationship between polysemy and type frequency. To redress this balance, an analysis was performed of seven samples from four languages, which can be classified in two sets. One set, consisting of English, Catalan and Mandarin Chinese, shows a monotonic distribution whereby type frequency decreases with increasing number of senses. The sole member of the other set is Greek, which exhibits a unimodal distribution with bisemy being more common than monosemy. In the English data, the ratio of monosemy to polysemy increases in favour of monosemy with increasing size of frequency-based samples. In line with Zipf, the principle of economy provides a good explanation of monotonicity. A cognitive account is offered of the unimodal distribution in Greek, which is argued to categorize less strongly and/or to use more fine-grained categories less often than the languages with a monotonic distribution do. This strategy boosts the usage frequency of more coarse-grained categories, which, in keeping with Zipf’s Law of Meaning Distribution, raises the incidence of polysemy.