Notes* 1. Thus Dixon (1972) makes the strong claim that "Each natural language is either strictly nominative-accusative, or strictly nominative-ergative in syntax (p. 129)." Clearly, our proposal involves a rejection of Dixon's claim, since we have found languages in which no "nominative" can be identified at all.
2. There is an apparent exception in Lisu; the nya morpheme which is the topic marker appears also in sentences involving no presuppositions at all, such as
In such sentences, nya is automatically added to the Agent if there is one, to the Dative if there is no agent, to the Object if there is no Dative, or to the Instrumental if there is no Object. Nya appears to function in two quite distinct ways, then, depending on whether or not the sentence contains any presupposed noun phrases. Hope (1974) refers to this distinction with his terms "primary" and "secondary" topicalization.
(i) swu nya atha do - a one topic marker knife forge - decl. marker "Someone is forging a knife." 3. Kuno (1972b:270) makes a similar point, using the term "theme" for what we are calling Topic: "The theme must be either anaphoric (i.e., previously mentioned) or generic, ..." The term "theme," incidentally, though used by Kuno in much the same way as we are using "topic," is not used by Prague school linguists in the same way. Firbas, for example, who has perhaps written more on the notions of "theme" and "theme" than anyone else, considers the "theme" to be the element which may be inferred from the context and which contributes the least to the development of the communication (Firbas 1966). Crucial to his development of this notion is the fact that every sentence can be said to have a theme. However, not every sentence has a topic.
4. So-called "dummy subjects" are only apparent exceptions, since they have no semantic content, and cannot therefore be included in a generalization concerning the semantic relationship between nouns and verbs.
5. We are not claiming, of course, that the topic has nothing to do with the comment. We do not expect to find a language in which the following type of sentence occurs:
(i) *My mother (topic), Harry can't stand chocolate mousse.However, we do not believe the oddness of a sentence like (i) to be a grammatical problem, but rather a pragmatic one. The point which we will continue to emphasize in the text is that the relationship between a subject and its verb is much more grammatically constrained than that between a topic and the comment. An explanation for this difference in terms of the functional role of topic and subject is offered in Section II(d) below.6. We are grateful to Paul Schachter for calling this fact to our attention. See also E.L. Keenan (Definition of Subject, this volume)
7. Talmy Givon (this volume) takes a different position. According to him, all agreement is topic-agreement, where "agreement" includes resumptive subject pronouns, as in(i) My mother she just came in.While we are convinced by the evidence supporting Givon's claim that agreement arises diachronically from resumptive pronouns, we do take issue with the claim that agreement morphemes and subject pronouns are indistinguishable synchronically. We know of no evidence against the hypothesis that by the time the pronoun has become a bound agreement morpheme the noun with which it agrees is no longer simply a topic but is integrated into the grammatical system as a subject. Evidence is, however, difficult to find, since none of the topic-prominent languages that we have investigated have well-developed anaphoric pronoun systems.8. We are grateful to Irwin Howard for calling to our attention the correlation between topic-prominence and special passives. For some discussion, see Howard (1969) and Kuno (1973). There is, however, some evidence that both Mandarin and Japanese are moving away from passives expressing only adverse effects, which may be due to the influence of English translations (G. Bedell, personal communication; see also Howard, 1969).
9. There are more serious objections to certain details of Teng's analysis. For example, Teng's sentential predicate is claimed to be a VP. But, as has been pointed out by Charles Tang and Marie-Claude Paris, the "sentential predicate" only obeys the tests for VP-hood (which exploit the pre-VP position of models and certain adverbs) if it is a lexicalized unit of some kind. Thus the "sentential predicate" tou teng "head ache" in 25 behaves like a VP, but yezi da "leaves big" in 23 does not:
For more discussion of "double-subject" sentences, see Li and Thompson (in preparation).
(i) Ta you tou teng le he again head ache aspect "He has a head-ache again."
(ii) *Ned - ke shu you yezi da le that - class. tree again leaves big aspect "That tree again has big leaves." 10. We are aware that subject markers (ga and ka) sometimes occur with the topics in "double subject" constructions in Japanese and Korean respectively. It appears that subject markers can appear with topics in certain discourse situations, but we do not at present understand this phenomenon well enough to be more precise.
11. An exception may be the deletion of a first-person pronoun on identity with a non-topic first person pronoun in a preceding clause. Thus, some speakers of Mandarin accept the following sentence, where the deleted constituent is understood to refer to the subject "I," and not to the topic "this character":
12. It is appropriate here to point out that there is no inconsistency between (1) using topic-predicate incompatibility as evidence in arguing that topics control deletion and (2) pointing out that topics are independent of predicated (see Section II(b) and II(c)). Topics may be independent of predicated in terms of grammatical processes and selectional relations if there is also a subject in the clause, but in a clause in which there is no subject distinct from the topic, as in the second clauses of 27-29, then of course subject and topic properties cannot be distinguished. The point is that the NP which is understood as filling the blanks in 27-29 is the topic of the first clause and not its subject. (We are grateful to Greg Lee and Fu Yi-Chin for their comments on this point.)
(i) Zheige zi wo bu renshi, suoyi bu ken zheiben xiaoshuo this character I not recognize so not read this novel "This character, I don't recognize, so (I) won't read this novel." 13. Following Soemarmo's (1970) analysis of Indonesian, we consider the topic-comment structure to be the one that involves the "left dislocation" of a constituent. Those sentences that contain an initial noun phrase marked by the postposition jang are analyzed as having the focus presupposition structure rather than the topic-comment structure. For example, the sentence
has a focus anak itu marked by jang and a presupposition: "someone bought shoes."
Anak itu jang membeli sepatu child Art. buy shoe "It was the child who bought shoes." 14. Of course, one could try to suggest that sentences such as 64 and 65 are derived from sentences with the object in final position by a rule of object proposing. It is important to note, however, that object proposing generally does not entail agent-deletion, which one would have to postulate to account for sentences such as 64 and 65. In addition, no such rule can be invoked to account for examples such as 60-63, which do not differ from the "pseudo-passive" in any way, except that the topic in the "pseudo-passive" has a selectional relationship with the verb.
15. See Hooper and Thompson (1973) for a discussion of this notion.
16. This formulation was suggested to us by Wallace Chafe (personal communication). For a discussion of case frames, see Fillmore (1968).
*Charles N. Li & Sandra A. Thompson. 1976. "Subject and Topic: A New Typology of Language." In: Charles N. Li (ed.). Subject and Topic. London /New York: Academic Press, pp. 486-489.