Quantifiers that are more monotone are easier to learn

Abstract

Linguistic universals have been hypothesized to bound the observed properties of natural languages at all levels of linguistic analysis (Greenberg 1966; Croft 2003; van der Hulst 2008). In the domain of semantics, such universals are often explained by appeal to general information-theoretic and cognitive facts, from efficient communication to ease of learning. This paper expands the range of the ease of learning explanation through a case study of monotone quantifiers. While most existing studies compare expressions which either do or do not satisfy a given universal property, we here define a metric for measuring monotonicity in degrees. A computational experiment shows that this degree correlates well with ease of learning, providing a finer-grained explanation in the ease of learning tradition.