* I am greatly indebted to Emily Bender, Charles Fillmore, Orin Gensler, Derek Herforth, Katsuya Kinjo, Kyoko Ohara, Masayoshi Shibatani, and the anonymous referees for the Journal of Pragmatics for their insightful comments and challenging questions.
1 This is a rather simplistic view. In fact, the alternative view that pragmatic inferences are necessary even for determining the propositional content has been gaining ground in recent years (Kempson 1986, Sperber and Wilson 1986, Levinson 1987, Carston 1988, Smith and Smith 1988, inter alia). However, the controversy is not directly relevant to the purpose of the present article, and accordingly I do not discuss this issue here.
2 The analyses advocated by these researchers are not formulated in terms of the Gricean theory of implicature. I nonetheless categorize them as falling under the implicature-only reductionist analysis because their primary stance is in principle to attribute the meanings of TE-linkage to the meanings of conjuncts. For example, Morita (1980: 313) notes that 'TE-linkage hardly ever conjoins clauses in a logical sequence or a temporal sequence; its meanings vary according to the meanings of the conjuncts' (translation mine).
3 For example, the 'gerund' of Archi, a Northeast Caucasian language, exhibits similar characteristics to those of TE-linkage (cf. Kibrik 1988).
4 For the traditional analysis of the 'verbal + TE' sequence, see Shibatani (1990: 227-28, 233-35).
5 Other examples of this type are: shii- 'force' + TE > shiite 'boldly/dare (do something)', hatas- 'accomplish' + TE > hatashite 'really', shitagaw- 'follow' + TE > shitagatte 'therefore'.
6 There are ten verbs that can serve as the auxiliary in this construction, e.g. ar- 'be located', k- 'come', ik- 'go', shimaw- 'put into an appropriate place', moraw- 'receive'. In classical Transformational Grammar treatments (e.g. Smith 1970, Nakau 1973, M. Inoue 1974), as well as in many current syntactic theories (e.g. McCawley and Momoi 1986, Shibatani 1987, Lee 1989, Sells 1990, Matsumoto 1990), the second verb is considered to be the main verb which takes a sentential or VP complement.
7 Minami (1974) categorizes Japanese connectives into three groups based on various co-occurrence restrictions: (A type) nagara, tsutsu 'while doing'; (B type) node 'because', temo 'although, even though', to 'and'; (C type) ga 'and, but', kara 'because', keredo 'but', shi 'and'. According to Minami, subtypes of te appear in all three categories.
8 On the basis of a corpus of 3,330 multi-predicate sentences sampled from various types of texts, Saeki (1975: 81) reports a total of 26 different lexical connectives (1,047 tokens altogether), of which te holds the foremost rank: it occurs 512 times, while the second most frequent connective, ga 'and/but', occurs only 141 times. According to K. Inoue (1983: 128-30), te appears most frequently in spontaneous speech (34.5% of all connectives) and in informal writing (27%). In formal writing such as newspaper editorials, te ranks second (17.2%) after ren'yoo linkage (36.9%). The actual occurrence of te is much more frequent than the numbers suggest because these data do not include cases in which the second predicate is a so-called auxiliary.
9 Lakoff (1971) claims that this statement also holds for the interpretation of coordination constructions (including and-linkage).
10 In order to express temporal sequence with (16), an adverbial such as sugu ni 'soon' or 5-fun-go ni '5 minutes later' must be inserted after the TE-predicate. That is, TE by itself does not implicate temporal sequence.
11 The notion of abduction, originally proposed by Charles S. Peirce, was introduced into linguistic circles by Henning Andersen (1973).
12 Pure abstract simultaneity is even more drastically incompatible with TE-linkage than is pure temporal sequence. For the latter, a perceived intention is sufficient to make the sentence acceptable in a temporal sequence reading, as shown above; whereas for the former, even when a perceived intention is present, e.g. watching TV and studying simultaneously, the sentence still cannot convey a simultaneous relation. This may be a remnant of a historical change: TE was originally an inflectional form of the old auxiliary verb tsu, which marked perfective aspect (Yamada 1954). Perfectivity of the TE-marked verb would lead easily and naturally to a temporal sequence relation between the two clauses, but not to a simultaneous relation.
13 Kuno claims that the sentences in (29) are ungrammatical. While they are certainly awkward, it is overly pedantic to call them ungrammatical. Sentences that violate the controllability constraint are in fact not uncommon.
14 It is important here to distinguish between wanting (e.g. I want to give you $5) and wishing (e.g. I want you to give me $5).
15 One might argue that there are some actions that a person may want to perform but would under no circumstances actually do, e.g. killing someone. However, if this is a genuine desire and not an idle wish, it must be supposed that the action is to some degree tempting to the person. In such a case, Donnellan argues, we will have to include a weakening of one's moral inhibitions as part of the set of circumstances under which one would be prepared to do the act.
16 Hart and Honoré, who have investigated causation in judicial contexts, also point out the distinction between causes and reasons, They note that 'a voluntary human action intended to bring about what in fact happens, and in the manner in which it happens, has a special place in causal inquiries ... when the question is how far back a cause shall be traced through a number of intervening causes, such a voluntary action very often is regarded both as a limit and also as still the cause even though other later abnormal occurrences are recognized as causes' (1959:39). For example,
If unusual quantities of arsenic are found in a dead man's body, this is up to a point an explanation of his death and so the cause of it: but we usually press for a further and more satisfying explanation and may find that someone deliberately put arsenic in the victim's food. This is a fuller explanation in terms of human agency; and ... we speak of the poisoner's action as the cause of the death; though we do not withdraw the title of cause from the presence of arsenic in the body -- this is now thought of as the "mere way" in which the poisoner produced the effect. Once we have reached this point ... we have something which has a special finality at the level of common sense: for though we may look for and find an explanation of why the poisoner did what he did in terms of motives like greed or revenge, we do not regard his motive ... as the cause of the death ... We do not trace the cause through the deliberate act. (ibid., 39-40, emphasis in original.)
17 Kortmann considers that the unmarked temporal relationship between a present-participial free adjunct/absolute and the matrix clause is simultaneity/overlap. Here TE-linkage differs significantly from participial free adjuncts/absolutes. As mentioned above, TE-linkage is not compatible with the simultaneous relation.
18 This claim does not deny the possibility that some semantic relations are weighted according to their informativeness. Hasegawa (1992:226-27) reports that means is more informative than temporal sequence on such a scale, and that the former is selected whenever it accords with the interpreter's world knowledge.
...[the following sentence] was unanimously translated as A by six [bilingual] native speakers of Japanese ...gomukan o pintikokku de tomete hi o kesu.
A: 'By pinching the rubber tube with a pinch-cock, extinguish the flame.'
B: 'Pinch the rubber tube with a pinch-cock, and extinguish the flame.'[This sentence], taken from a high-school science textbook, is part of an experiment procedure. In this experiment, two rubber tubes are used: one connecting a gas pipe and a burner, and the other a flask and a glass tube. After careful consideration, [the participants in the project] recognized that A instructs the students to turn off the gas flame by an unusual and dangerous means, a means which should not appear in a science textbook. Thus [the participants] rejected A and took gomukan 'rubber tube' to be the one connecting the flask and the glass tube ... Note that when [they] first translated [the sentence], [they] were totally aware of these linguistic and extralinguistic contexts, and yet employed the unmarked parsing strategy ... because pinching the tube can extinguish the flame.
19 The enablement relation is similar to, but weaker than, the cause relation. The cause relation indicates both necessary and sufficient conditions, whereas the enablement relation indicates only the former.