This paper offers explicit formulations of the processes underlying the notion of the grammaticization of topics into subjects, which has been widely alluded to in the literature (Li & Thompson 1976, Givon 1979, Plank 1979, Mallinson & Blake 1981). The scope of our investigation covers several typological groups of languages. Because of the need to come to grips with the nature of a grammatical topic, we first look at Japanese. English represents those language in which the topic and subject functions have converged on an agentive nominal (of a transitive clause), whereas the deep ergative language Dyirbal represents those in which the two functions have largely converged on a patient nominal. In other words, English is an instance of those languages in which "the nominative marks a grammaticalized topic," and Dyirbal those in which "the absolutive represents a grammaticalized topic" (Mallinson & Blake 1981: 109). Philippine languages, on the one hand, offer a transition from those languages in whieh a topic is optional (e. g. Japanese) and those in which a topic is made an obligatory constituent of major clause types and has become a subject to a great extent; on the other hand, they bridge a gap between accusative languages and ergative languages. Philippine languages and ergative languages together form a group of languages that offer a challenge for an account on the rise of split syntax.
The present paper thus deals with the following problems: 1) definitions of the grammatical topic and of the grammatical subject, 2) clarification of the notion of topic-prominent vs. subject-prominent languages (Li & Thompson 1976), 3) explication of the status of the Philippine-style topic, and 4) explanation for the rise and nature of split syntax.