Gender, number and person: a three-way interaction

According to the principle of informativity, more useful information is more likely to be coded than less useful information. This principle underlies Greenberg’s Universals 37 and 45, which state that a larger number of (sex-based) gender contrasts are on average found in singular than in non-singular forms. As an increase in group size lowers the probability of same sex groups, gender-specific personal pronouns are less useful in the non-singular than in the singular. Curiously enough, the principle of informativity also makes the opposite prediction: in the first person, non-singular forms are predicted to show a gender contrast more frequently than singular forms do. Moreover, exclusive forms are predicted to develop gender distinctions more often than inclusive forms. In this paper, these predictions are put to a typological test. A total of 51 languages from 17 different families and 24 different genera has been found in which the first person singular personal pronoun is gender-neutral while one or more of its non-singular counterparts inflect for gender. It is argued that Universals 37 and 45, as well as their counterclaim, are empirically adequate. This apparent paradox dissolves in a three-way interaction of gender, number and person in pronominal paradigms. In keeping with the principle of informativity, there is a bias in favour of gender marking in the third person singular but also a bias in favour of gender marking in the first person non-singular.