The Japanese particles 'wa' and 'ga': A quantitative study of a text.

Abstract

This is a study of the distribution of the Japanese particles wa and ga as markers of subject. One of the problems in the theories that have been developed so far is that they are either circular or based on concepts which are not well defined. Another problem is that the examples used in these studies are invented by researchers for the explanation of their theories. But numerous instances of uses of the particles which have not been even considered so far can be found in natural data. This is a quantitative study of more than 2,000 subjects which are marked by wa or ga in a two hundred page text. Each token is assigned values for 15 factors which cause a subject to be marked more often by wa or ga. All 15 factors are defined as objectively as possible, and no vaguely defined concepts (e.g. theme) are used in the definition. The statistical techniques of logistic regression and Chi-square test, are used to examine the statistical significance of the frequencies of the tokens gained from the observation of the data. It will be shown that about 70% of all the instances of wa and ga in the text can be explained by 10 statistically significant factors. Among these, 6 are strong factors, which favor the use of wa or ga more than 95% of the times that they are present. The factor Second member of a contrastive pair favors wa, while the other 5 factors (Idiomatic Expressions, Intra-NP Position, Double Subject Construction, Existential/Presentative Verbs, and Focus) all favor ga. There are 4 weak factors. If a subject is not assigned any of the strong factors, and if it is not in a subordinate clause, these factors favor the use of wa or ga about 90% or more of the times that they are present. The factor Quasi-Passive-Verb favors ga, while three factors (Sentence-Initial Position, Multiple Predicate Construction, and Referential Distance) favor wa.