(vii) Assuming an affirmative answer to (iii), should the analysis rely on the
discourse ...... (26) John-wa [[[Mary-ga teeburu-no ue-ni oite-oi-ta] ringo] to. John-
Top ...
Japanese IHRCs: Towards an improved understanding of their syntactic, semantic and pragmatic properties* Alexander Grosu Koji Hoshi
Tel Aviv University Keio University
1. Introduction The internally-headed relative constructions (henceforth: IHRCs) of Japanese have benefited from the analytical attention of a sizable number of linguists (many of whom are native speakers of the language), possibly more so than the IHRCs of any other language, as far as we can tell. One result of this state of affairs is that the descriptive properties of these constructions have been investigated in a considerable degree of detail. Another result is, however, that there is a great deal of disagreement among earlier researchers concerning the exact nature of these properties, and – not unrelatedly – their optimal analysis. Insofar as descriptive properties are concerned, disagreement has emerged concerning (at least) the following issues: (i) Do IHRCs exist at all? (ii) Are IHRCs sensitive to syntactic islands? (iii) Are IHRCs always definite descriptions? (iv) Are definite referential internal heads (henceforth: IHs) infelicitous? (v) Can IHs be syntactically unrepresented? Concerning analytical issues, two of them that have prominently come up are: (vi) Assuming a positive answer to (ii), can the island-sensitive unbounded dependency originate within the IH itself? (vii) Assuming an affirmative answer to (iii), should the analysis rely on the discourse mechanism of E-type anaphora, or should a grammatical mechanism with partly comparable semantic properties be used instead? The inter-related goals of this paper are to provide novel empirical arguments in support of affirmative answers to (i)-(v) and a negative answer to (vi), and to strengthen thereby the empirical basis of the analysis of Japanese IHRCs in Grosu & Landman (2012), as refined in Landman (ms.), which relies on grammatical, not discourse, mechanisms (see (vii)). In so doing, we will also consider, discuss, and refute earlier challenges to the theses just enunciated.
2 2. Homophony of IHRCs with adverbials Mihara (1994) suggested that IHRCs do not exist and that all presumed IHRCs are in fact sentential adverbials (see also, e.g., Murasugi 1994 and Hoshi 1996). This suggestion relied on the widely recognized fact that some IHRCs are string-wise homophonous with adverbial clauses. However, it was subsequently pointed out that this view is too strong, and that genuine IHRCs do exist (see, e.g., Kuroda 1999). Before providing evidence to this effect, we note that homophony arises due to the fact that the suffix –no, which has many functions in Japanese, but functions as a semantically empty nominalizer in IHRCs, typically bearing a Case marker in such cases, sometimes gives rise to affixal sequences homophonous with adverbial markers. In particular, the sequences -no-ga, -no-o, -no-ni, which consist of the nominalizer followed, respectively, by Nominative, Accusative, and Dative Case markers, may also serve as adverbial markers (in traditional grammar terminology: 'subordinating conjunctions'). For completeness, we note that the adverbial marker –no-ni, which determines a temporal simultaneity construal of the subordinate clause and the matrix, is also homophonous with the adversative adverbial marker –noni, which has a clear status as a subordinating conjunction in Japanese. On the assumption, which we will argue for below, that both genuine IHRCs and homophonous adverbial constructions exist, we illustrate the ambiguity of three kinds of sequences in (1)-(3) respectively.1 Observe that while the IHRCs and the adverbials in these examples are string-wise homophonous, we do not assign them completely identical structural representations. In particular, we assume that while IHRCs occur in argument positions within the matrix, homophonous adverbials do not, the corresponding argument position being occupied by a null pronoun (= pro) which is anaphoric to the boldfaced nominal within the subordinate clause. (1) a. [[Daidokoro-no mado-kara siroi neko-ga haitte-kita]-no]-ga kitchen–Gen window-from white cat-Nom come.in-Past-NML-Nom sakana-o totte nige-ta. IHRC fish-Acc taking run-Past 'A white cat came in from the kitchen window and it stole a fish and ran away.’ b. [[Daidokoro-no mado-kara siroi neko-ga haitte-kita]-no]-ga pro kitchen–Gen window-from white cat-Nom come.in-Past-NML-Nom sakana-o totte nige-ta. Adverbial fish-Acc taking run-Past 'As a white cat came in from the kitchen window, it stole a fish and ran away.’ Or: ‘A white cat having come in from the kitchen window, it stole a fish and ran away.’ (2) a. Anthony-wa [[doroboo-ga huta-ri nige-teiru]-no]-o tukamae-ta. IHRC Anthony-Top thief-Nom two-Cl run.away-Prog-NML-Acc catch-Past ’Two thieves were running away, and Anthony caught them.’ 1
In general, IHRCs with a Dative Case marker are judged relatively a bit degraded in comparison with ones with Nominative and Accusative markers for yet unknown reasons. Following Kuroda (1999), however, we assume that they can be generated by the grammar of Japanese. 2
3 b. Anthony-wa[[doroboo-ga huta-ri nige-teiru]-no]-o pro tukamae-ta. Adverbial Anthony-Top thief-Nom two-Cl run.away-Prog-NML-Acc catch-Past ’As two thieves were running away, Anthony caught them.’ Or: ‘Two thieves being on the run, Anthony caught them.’ (3) a. John-wa [[Mary-ga kesa sono kado-ni kuruma-o John-Top Mary-Nom this morning that corner-at car-Acc tome-ta]-no]-ni (gogo) butukat-ta. park-Past-NML-Dat (afternoon) bump.into-Past ‘Mary parked a car at that corner this morning, and John bumped into it (this afternoon).’ IHRC b. John-wa [[Mary-ga kesa sono kado-ni kuruma-o John-Top Mary-Nom this morning that corner-at car-Acc tometa]-no]-ni (#gogo) pro butukat-ta. park-Past-NML-Dat (afternoon) bump.into-Past ‘The moment Mary parked a car at that corner this morning, John bumped into it (#this afternoon).’ Temporal Simultaneity Adverbial c. John-wa [Mary-ga kesa sono kado-ni kuruma-o John-Top Mary-Nom this morning that corner-at car-Acc tometa]-noni pro butukaranakat-ta. park-Past-although bump.into-Neg-Past ‘Although Mary parked a car at that corner this morning, John did not bump into it.’ Adversative Adverbial Observe that while the IHRC and the corresponding adverbial constructions are not synonymous, the difference in meaning between them is not always striking, and it is in principle possible to confuse one with the other in certain cases, such as (1)-(2) and (3ab) without gogo 'this afternoon.' At the same time, confusion is highly unlikely to occur in connection with the adversative adverbial reading in (3c), and it is, of course, totally excluded in the full version of (3b), which is semantically incoherent. Be this as it may, it is important to note that confusion has arguably occurred in a number of earlier studies, giving rise to incorrect conclusions about the properties of IHRCs, as will be seen in subsequent sections of this paper. Returning now to Mihara's (1994) suggestion, there are (at least) two sets of facts that argue against it. One fact, already noted in the earlier literature (e.g., Watanabe 1992, Hoshi 1995, Kuroda 1999), is that there is no adverbial marker synonymous with the sequence -no-no, in which the second token of -no is a marker of Genitive Case. Accordingly, any acceptable relevant construction ending with this sequence must be viewed as either a genuine IHRC, or as a nominalized sentential complement bearing Genitive Case, a possibility that also exists (see (22b) and the paragraph which precedes example (20)). A second type of fact, which, to the best of our knowledge, has not been noted in the earlier literature, is the existence of a variant of IHRCs that we may call 'Split Headed Relative Constructions' (SHRCs). The term 'split' purports to characterize the fact that a numeral which may in principle occur within an IH, may also occur outside CP, in particular, immediately to the right of the Case marker. Such SHRCs are fully 3
4 synonymous with minimally different IHRCs, but may not be construed as adverbials. A minimal pair of synonymous IHRCs and SHRCs is provided in (4). (4) a. John-wa [[Mary-ga nempai-no happyoosya-o huta-ri kuukoo-de IHRC John-Top Mary-Nom elderly-Gen speaker-Acc two-Cl airport-at mat-asete-oi-ta]-no]-o hoteru-e tureteit-ta. wait-Caus-Aux-Past-NML-Acc hotel-to take-Past 'Mary had two elderly speakers waiting at the airport, and John took them to a hotel.' b. John-wa [[Mary-ga nempai-no happyoosya-o kuukoo-de SHRC John-Top Mary-Nom elderly-Gen speaker-Acc airport-at mat-asete-oi-ta]-no]-o huta-ri hoteru-e tureteit-ta. wait-Caus-Aux-Past-NML-Acc two-Cl hotel-to take-Past 'Mary had two elderly speakers waiting at the airport, and John took them to a hotel.' We note that in order to get a reading of (4b) that is synonymous with (4a), one must not pause before the numeral, because such a pause favors the partitive reading 'Mary had elderly speakers waiting at the airport, and John took two of them to a hotel', on which the left sister of the numeral is an IHRC that serves as partitive complement. Keeping this in mind, we wish to emphasize that (4a) is string-wise homophonous with an adverbial construction translatable as "as Mary had two elderly speakers waiting at the airport, John took them to a hotel", and that the non-partitive reading of (4b) is not. 3. The island-sensitivity of IHRCs The question whether IHRCs are sensitive or insensitive to syntactic islands has been a highly controversial one in the earlier literature. We propose to argue in this section that IHRCs are sensitive to at least two types of island constraints, in particular, to the Complex NP Constraint (Henceforth: CNPC) and the Adjunct Island Constraint (henceforth: AIC), and that earlier claims to the contrary have relied on data that allow adverbial construals. Given the possibility of such a construal, island insensitivity is unsurprising, because the adverbials at issue exhibit discourse anaphoric dependencies, which are notoriously insensitive to island constraints. Concerning the CNPC, Watanabe (1992, 2003) proposed, on the basis of contrasts like that in (5a-b), that the IH may be part of an unbounded dependency, but not in violation of the CNPC. (5) a. Mary-ga [[[John-ga [zibun-no gakusei-ga zyuuyoona kasetu-o Mary-Nom John-Nom self-Gen student-Nom important hypothesis-Acc teian-si-ta to] zimansite-ita]-no]-no kekkan]-o siteki-si-ta. propose-do-Past Comp boasted-had-NML-Gen defect-Acc point.out-do-Past ‘John had boasted that his student proposed an important hypothesis and Mary pointed out a defect in it.’ b. *Mary-ga [[[John-ga [[atarasii kasetu-o teiansita] gakusei]-o Mary-Nom John-Nom new hypothesis-Acc propose-Past student-Acc homete-ita]-no]-no kekkan]-o sitekisita. praise-had-NML-Gen defect-Acc pointe.out-do-Past 4
5 ‘John praised the student [who proposed a new hypothesis] and Mary pointed out a defect in it.' c. [[Kyoozyu-ga [[sono daigakuinsei-ga [e] kaita] ronbun]-o professor-Nom that grad-student-Nom write-Past paper-Acc homete-i-ta]-no]-ga kondo zyosyu-de saiyoo-sare-ru koto-ni-nat-ta. praise-Prog-Past-NML-Nom now instructor-as hire-Pass-Pres come.to.be-Past 'That graduate student of whom the professor praised the paper he had written has been appointed as an instructor.' Kitagawa (2005), while agreeing with Watanabe's judgments about (5a-b), attributes the deviance of (5b) to a violation of a pragmatic constraint, which has been dubbed the 'Relevancy Condition' in Kuroda (1975-76), and offers the acceptable example in (5c) as evidence that the CNPC may be violated in Japanese IHRCs with impunity. Kuroda's original formulation of this condition consisted of the two statements in (6a-b); in section 7, we introduce and justify a refined version of (6). (6) a. The Relevancy Condition For an internally headed relative to be acceptable, it is necessary that it be interpreted pragmatically in such a way as to be directly relevant to the pragmatic content of its matrix clause. b. Sub-condition The two events represented by the internally headed relative and the matrix clause involve the same temporal interval and the same location. What Kitagawa fails to notice is that the constituent in (5c) within the most inclusive set of brackets and marked with -ga is homophonous with an adverbial, so that this example is translatable as in (7). (7) The professor having praised the paper that that graduate student had written, he (= the student) was appointed as an instructor. Adverbial We will now show that if an IHRC interpretation is coerced by means of either one of the tests indicated in section 2, the violation of the CNPC results in unacceptability. Consider the unacceptability of the (b) sub-case of (8). (8) a. [[Kyoozyu-ga [[huta-ri-no daigakuinsei-ga [e] kaita] ronbun]-o professor-Nom two-Cl-Gen grad-student-Nom write-Past paper-Acc homete-i-ta]-no]-ga kondo zyosyu-de saiyoo-sare-ru koto-ni-nat-ta. praise-Prog-Past-NML-Nom now instructor-as hire-Pass-Pres come.to.be-Past 'The professor having praised the papers that two graduate students had written, they (= the students) were appointed as instructors.' b. *[[Kyoozyu-ga [[daigakuinsei-ga [e] kaita] ronbun]-o professor-Nom grad-student-Nom write-Past paper-Acc homete-i-ta]-no]-ga huta-ri kondo zyosyu-de saiyoo-sare-ru praise-Prog-Past-NML-Nom two-Cl now instructor-as hire-Pass-Pres 5
6 koto-ni-nat-ta. come.to.be-Past 'The professor praised the papers that two graduate students had written, and they (= the students) were appointed as instructors.' [intended reading] (8b) shows the unacceptability of the SHRC counterpart of (8a), a slightly adapted version of (5c); in the (b) sub-case, an adverbial construal is excluded, and sensitivity to the CNPC is exposed. As for Kitagawa's suggestion that Watanabe's (5b) runs afoul of the Relevancy Condition, he provides no justification for this statement, and we do not see why this should be so. We provide in (9) a slightly modified variant, which is unquestioningly consistent with the Relevancy Condition (for reasons we will make explicit in section 7), and which is equally unacceptable. Note that the constituent within the second most inclusive pair of brackets bears the Genitive Case, and thus excludes an adverbial interpretation. (9) *Mary-ga [[[John-ga [EHRC [ [e] atarasii kasetu-o teiansi-ta] gakusei]-o Mary-Nom John-Nom new hypothesis-Acc propose-Past student-Acc hidoku homete-ita]-no]-no akirakana kekkan]-o suguni siteki-si-ta. extravagantly praise-had-NML-Gen obvious defect-Acc promptly point.out-do-Past ‘John extravagantly praised the student [who had proposed a new hypothesis] and Mary promptly pointed out an obvious defect in it.' Turning now to the AIC, Hoshi (1995) and Kuroda (1999) were led to the conclusion that IHRCs are insensitive to it by data like (10). Observe, however, that this example has not only the presumed IHRC reading, but also the indicated adverbial reading. (10) John-wa [[Mary-ga [kaseehu-san-ga ringo-o teeburu-no ue-ni oi-ta] ato-de John-Top Mary-Nom housemaid-Pol-Nom apple-Acc table-Gen on put-Past after dekaketesimat-ta]-no]-o (pro) kossori totte tabetesimat-ta. leave.home-Past-NML-Acc secretly pick.up and eat.up-Past ‘Mary left home after her housemaid put apples on the table and John surreptitiously picked them up and ate them up.’ IHRC ‘As Mary left home after her housemaid put apples on the table, John surreptitiously picked them up and ate them up.’ Adverbial When the adverbial reading is excluded, the sensitivity of IHRCs to this constraint is laid bare. We show this in (11)-(12) with the two tests proposed in section 2, in particular, the Genitive Case test and the SHRC test respectively. (11)?*Mary-wa [[[John-ga [kare-no gakusei-ga atarasii kasetu-o Mary-Top John-Nom his student-Nom new hypothesis-Acc teiansi-ta]-node kanki-no koe-o age-ta]-no]-no propose-Past because joy-Gen voice-Acc raise-Past-NML-Gen 6
7 akirakana kekkan]-o suguni sitekisi-ta. obvious defect-Acc promptly point.out-Past `John shouted for joy because his student proposed a new hypothesis, and Mary promptly pointed out an obvious defect in it.’ (12) *? John-wa [[Mary-ga [imooto-ga nempai-no happyoosya-o kuukoo-de John-Top Mary-Nom sister-Nom elderly-Gen speaker-Acc airport-at mat-asete-oi-ta]-node totemo otituk-anakat-ta]-no]-o huta-ri wait-Caus-Aux-Past-because very stay.calm-Neg-Past-NML-Acc two-Cl hoteru-e tureteit-ta. hotel-to take-Past ‘Mary was very nervous because her sister had two elderly speakers waiting at the airport, and John took them to a hotel.’
4. The principal semantic and pragmatic properties of Japanese IHRCs The earlier literature has pointed out three properties, two semantic and one pragmatic, some, and in some cases all, of which distinguish the IHRCs of Japanese from those of many other languages, as well as from restrictive EHRCs and from discourses with anaphoric dependencies. The semantic properties are [A] and [B]. The pragmatic property is [C]: [A] The IH is an indefinite or quantified DP with relative-internal scope (Hoshi 1995 and Shimoyama 1999, 2001). [B] The entire IHRC is necessarily construed as a definite description (Hoshi 1995 and Shimoyama 1999, 2001). [C] The entire IHRC is subject to the Relevancy Condition (Kuroda 1975-76), which was stated in its initial form in (6), and which, as promised earlier, will be restated in a refined form in section 7. [A] distinguishes Japanese IHRCs from the IHRCs of such languages as Cuzco Quechua and Navajo, where strongly quantified IHs take matrix scope (see, e.g., Hastings 2004 and Grosu 2012 respectively), as well as from the EHRCs of Japanese, whose quantified external heads always have matrix scope (see Hoshi 1995 and Shimoyama 1999, 2001). We note for completeness that while the scope of an IH needs to be internal to the relative clause, it need not be maximally narrow, as illustrated in (13) (due to Shimoyama, p.c., cited in Erlewine and Gould 2012), where the IH takes scope over relative-internal negation. (13) [[Nani-ka muzukasii hon-o dono kyoozyu-mo yoma-nakat-ta]-no]-o what–ka difficult book-Acc which prof.-mo read-Neg-Past-NML-Acc ano gakusee-wa yon-deiru-rasii. that student-Top read-Prog-Evid ‘Apparently, there is some difficult book or other that no professor has read, and that student is reading it.’ 7
8
[B] distinguishes the IHRCs of Japanese from those of languages like Lakhota and Navajo, which belong to the restrictive relative class, and thus allow the full range of quantificational forces available in the language, and in particular, existential force (for details, see, e.g., Grosu 2012 and references therein). Property [B] was widely recognized in some earlier literature, although specific writers proposed different analytical implementations of this effect, e.g., Hoshi (1995) and Shimoyama (1999, 2001) proposed accounts that relied on E-type anaphora, while Grosu (2010), Grosu & Landman (2012), Landman (ms.) proposed accounts based on a relative-internal grammatical mechanism (to which we return in section 7). [C] distinguishes Japanese IHRCs from EHRCs (Kuroda 1975-76), from discourses with anaphoric dependencies, and from the IHRCs of many other languages, including some that possess properties [A]-[B], in particular, Cuzco and Imbabura Quechua (Hastings 2004, p.c.). As far as we know, the only other language for which Relevancy Condition effects have been reported in earlier literature is Korean (see, e.g., Kim 2007, 2008 and references therein). In the remainder of this section, we will consider a challenge by Kubota & Smith (2007) to the view that Japanese IHRCs exhibit property [B]. These writers bring up contrasting data like those in (14a-b), and urge on this basis the conclusion that Japanese IHRCs may be either definite or indefinite, depending on the pragmatic context. We reproduce these data with the translations assigned to them by Kubota and Smith, with which we disagree (see below for what we view as the correct translations). (14) a. (At the security check of an airport:) Dono zyookyaku-mo [[pro poketto-ni koin-ga haittei-ta]-no]-o every passenger pocket-Dat coin-Nom in.be-Past-NML-Acc toridasi-te torei-ni nose-ta. pick.up tray-Dat put-Past ‘Every passenger picked up the coins that s(he) had in (his/her) pocket and put them on the tray.’ b. (At the ticket gate of a train station:) Dono zyookyaku-mo [[pro saihu-ni kaisuuken-ga haittei-ta]-no]-o every passenger wallet-Dat coupon.ticket-Nom in.be-Past-NML-Acc toridasi-te kaisatu-ni ire-ta. pick.up ticket.checker-Dat put-Past ‘Every passenger picked up a coupon ticket that s(he) had in (his/her) wallet and put it in the ticket checker.’ In their view, this conclusion is justified by the observation that (14a) is most naturally understood as implicating that every passenger put on the tray all the coins there were in his/her pocket, while (14b) is most naturally understood as implicating that every passenger put a single ticket in the checker, while allowing for the possibility that there were multiple tickets in the wallets of some or all of the passengers. In the authors' view, the pragmatic assumptions associated with the two matrices, i.e., that a passenger is expected to put all the coins in his/her pocket on the tray, but only one ticket into the
8
9 checker, apparently coerce a definite construal of the IHRC in (14a) and an indefinite construal of the one in (14b). We do not think, however, that the pragmatic factors just noted affect the definiteness of the IHRCs, nor the IHs insofar as definiteness or indefiniteness is concerned. We do, however, think that they affect the interpretation of the IHs in ways to which we turn directly. Thus, consider (14'), in which we provide what we view as the correct translations of the data in (14). (14') a. (At the security check of an airport:) Dono zyookyaku-mo [[pro poketto-ni koin-ga haittei-ta]-no]-o every passenger pocket-Dat coin-Nom in.be-Past-NML-Acc toridasi-te torei-ni nose-ta. pick.up tray-Dat put-Past ‘For every passenger x, x had coins in x's pocket and x picked them up and put them on the tray.’ b. (At the ticket gate of a train station:) Dono zyookyaku-mo [[pro saihu-ni kaisuuken-ga haittei-ta]-no]-o every passenger wallet-Dat coupon.ticket-Nom in.be-Past-NML-Acc toridasi-te kaisatu-ni ire-ta. pick.up ticket.checker-Dat put-Past ‘For every passenger x, x had a coupon ticket in x's wallet and x picked it up and put it in the ticket checker.’ Observe first that we translated the IH with a plural expression in (14'a) and with a singular one in (14'b). We feel free to do this, because bare nouns in Japanese are unmarked for the singular/plural distinction. Second, we translated both IHs with indefinite expressions in English, because we know of no reason to do otherwise. Third, observe that the translations are discourses that consist of two coordinate sentences related by a cross-sentential E-type anaphoric dependency. Fourth, we will assume the following points, which anticipate the more detailed presentation of the semantics and pragmatics of IHRCs that will be outlined in section 7: The actual semantics for IHRCs, assuming that felicity and grammaticality conditions are satisfied, is not unlike the semantics that is usually proposed for minimally different discourse anaphora cases. Furthermore, while the Relevancy Condition (which will be substantially refined in section 7) is a felicity condition on IHRCs, it is not a felicity condition on discourses. Nonetheless, nothing prevents discourses from also satisfying this condition, and when this happens, the Relevancy Condition may achieve a pragmatic effect by increasing the coherence of a discourse, because the bits are linked via coherent events. Most importantly, we will argue in section 7 (and will assume here) that a crucial consequence of the Relevancy Condition is that the matrix of the IHRC constrains the relative. We thus assume that when a minimally different discourse obeys the Relevancy Condition, the second sentence in the discourse constrains the interpretation of the first. Keeping in mind what was said in the preceding paragraph, we propose that the distinct pragmatic assumptions associated with the matrices of (14'a) and (14'b) take advantage of the indeterminacy for Number of the corresponding IHs to trigger the 9
10 plural/singular construals indicated in the corresponding English translations. Fred Landman (p.c.) kindly pointed out to us that indeterminacy of a slightly different sort can be exploited in a comparable way in English conditional constructions like (15). (15) a. (At the security check of an airport) If you have a key with you, you have to put it on the tray. b. (A tenant who has lost his key says to another tenant:) If you have a key with you, please put it in the lock and open the door. It is well-known that indefinite expressions like those boldfaced in (15) may in principle exhibit (defeasible) numerical implicatures that trigger an 'exactly' construal (Kadmon 1990). In (15a), the numerical implicature potentially associable with a key is blocked by the pragmatics of the main clause, with the result that this sentence is naturally interpretable as "if you have at least one key in your pocket, you have to put (all) the keys in your pocket on the tray." In (15b), on the other hand, the pragmatics of the main clause block an 'at least' construal of a key, with the result that the implicature kicks in. Crucially, however, (15b) is fully compatible with a situation in which the addressee has more than one key that can open the door in his/her pocket, and correlatively, the English translation of (14'b) is also compatible with a situation in which the passengers at issue have more than one coupon-ticket in their wallets. The singularity of the boldfaced expressions is simply a consequence of the pragmatics associated with the main clause/second sentence in these examples, in particular, with the fact that a single key/coupon ticket out of possibly many is contextually relevant (see Kadmon 1990 for discussion of English data like there was once a doctor in London; he took good care of his patients, which does not necessarily imply that there couldn't have been additional doctors in London). What has just been shown extends straightforwardly to the Japanese example in (14'b), which is also compatible with situations in which the passengers had multiple coupontickets in their wallets. We conclude that the contrast in (14') does not show what Kubota and Smith thought it shows. Rather, given a particular (pragmatically constrained) interpretation of the IH, the IHRC simply picks out the maximal entity associated with that interpretation. Ergo, the view that property [B] is a property of Japanese IHRCs in general is in no way endangered by data like (14'). For the sake of completeness, we wish to point out that the pragmatics of the main clause of an IHRC cannot overcome quantitative information explicitly specified in the IH. To see this, consider the variant of (14'b) shown in (16). (16) (At the ticket gate of a train station:) Dono zyookyaku-mo [[pro saihu-ni nizyuu-mai-no kaisuuken-ga every passenger wallet-Dat twenty-Cl-Gen coupon.ticket-Nom haittei-ta]-no]-o toridasi-te kaisatu-ni ire-ta. in.be-Past-NML-Acc pick.up ticket.checker-Dat put-Past ‘For every passenger x, x had twenty coupon-tickets in his/her pocket and put them in the ticket checker.
10
11 This example, as well as its English translation, is only adequate in a situation where passengers have to use twenty coupon-tickets in order to get past the checker (and allows for the possibility that they may have more than twenty coupon-tickets in their wallets), but in a situation where a single coupon-ticket is sufficient to get past the checker, i.e., in the 'normal' one in our world, (16) is not open to the interpretation that every passenger put a single coupon-ticket in the checker.
5. Can definite referential DPs be felicitously used as IHs? We have so far brought up examples of IHRCs in which the IH is a quantified nominal. Can a definite referential DP, such as a proper name or a DP with a demonstrative determiner, also function as IH? From the perspective of proponents of an analysis that relies on E-type anaphora, this question can be re-formulated as follows: Is E-type anaphora the only mechanism for analyzing Japanese IHRCs, or can one also resort to referential anaphora? Shimoyama (2001, Chapter 3) noted contrasts like the one between (17a) and (17b), and proposed to view the deviance of (17b) as indicative of the fact that E-type anaphora is the only mechanism available to the IHRCs at issue. As for the acceptability of (17a), she suggested that the former is licensed by the possibility of an adverbial construal. (17) a. Taro-wa [[daidokoro-no mado-kara Lucky-ga haitteki-ta]-no]-o Taro-Top kitchen-Gen window-from Lucky-Nom come.in-Past-NML-Acc tukamae-ta. catch-Past ‘Taro caught Lucky, as she came in from the kitchen window.’ b. ?*Ken-wa [[[Naomi-ga ofisu-ni Lucky-o tureteki-ta]-no]-no ke]-o Ken-Top Naomi-Nom office-Dat Lucky-Acc bring-Past-NML-Gen hair-Acc kat-ta. cut-Past ‘Naomi brought Lucky to the office and Ken cut her hair.’ This proposal was challenged by Kitagawa (2005, section 5.1), who suggested that (17b) is degraded due to a failure to satisfy the Relevancy Condition. He offered no detailed explanation for how exactly this condition is violated, but offered the data in (18) in support of the conclusion that definite referential DPs are possible IHs (we reproduce Kitagawa’s data with his own English translations, with which we disagree). (18) a. [Soba-no nagaisu-no ue-de Lucky-ga nete-i-ta]-no]-ga Nearby-Gen couch-Gen on Lucky-Nom sleep-Prog-Past-NML-Nom ookina akubi-o si-ta. big yawn-Acc do-Past ‘Lucky, who was sleeping on a nearby couch, made a big yawn.’ b. [[[[[Lucky-ga atama dake dasite-i-ru]-no]-no usiro]-ni Lucky-Nom head only show-Prog-Pres-NML-Gen behind-at ututtei-ru]-no]-wa kimi ka? 11
12 appear-Pres-NML-Top you Q ‘Is it you in the photo standing behind Lucky, who is sticking out his head?’ We believe that Kitagawa's thesis is untenable for a number of reasons. Observe first that the bracketed constituent in (17b) bears Genitive Case, and is thus an unambiguous IHRC, in contrast to the bracketed constituent in (17a), which bears Accusative Case and is also construable as an adverbial (see section 3). Next, consider (17b'), which has been obtained from (17b) by replacing the referential IH with an existentially quantified one (see Shimoyama 1999, 2001). This example is acceptable, unlike (17b), and this makes it implausible that the latter violates the Relevancy Condition. Rather, the contrast between (17b) and (17b') points to the conclusion that the deviance of the former is due to the definite referential status of its IH, and that the contrast between (17a) and (17b) is due to the availability of an adverbial construal in the former, but not in the latter. The latter part of this conclusion concerns IHRCs with any kind of definite referential IHs, and thus also predicts the contrast in (17'), where the IHs are demonstrative expressions. (17) b'. Ken-wa [[[Naomi-ga ofisu-ni haiiro-no neko-o tureteki-ta]-no]-no Ken-Top Naomi-Nom office-Dat gray-Gen cat-Acc bring-Past-NML-Gen ke]-o kat-ta. hair-Acc cut-Past ‘Naomi brought a gray cat to the office and Ken cut her hair.’ (17’) a. Taro-wa [[daidokoro-no mado-kara sono neko-ga haitteki-ta]-no]-o Taro-Top kitchen-Gen window-from that cat-Nom come.in-Past-NML-Acc tukamae-ta. catch-Past ‘Taro caught that cat, as she came in from the kitchen window.’ b. ?*Ken-wa [[[Naomi-ga ofisu-ni sono neko-o tureteki-ta]-no]-no ke]-o Ken-Top Naomi-Nom office-Dat that cat-Acc bring-Past-NML-Gen hair-Acc kat-ta. cut-Past ‘Naomi brought that cat to the office and Ken cut her hair.’ As for Kitagawa's data in (18), they do not support his thesis, because the relevant constituents in each sub-case allows a homophonous non-IHRC construal. Thus, the constituent in (18a), which bears Nominative Case, is homophonous with an adverbial, and the one in (18b) is homophonous with a nominalized sentential complement, to which a Genitive Case-marker is attached. We note that IHRCs are sometimes homophonous not only with adverbials, but also with complements, a widely recognized fact in the earlier literature (and, we may add, one that seems to be extremely common cross-linguistically; Culy 1990 in fact suggests that the IHRCs of all languages are homophonous with complements). To illustrate, consider (19), where the relevant constituent is ambiguous between the two interpretations at issue, as well as (20), which shows that careful selection of matrix predicates may coerce one of the two interpretations. We thus submit that the correct translations of Kitagawa's data in (18) are not the ones he provided, but rather the ones shown in (21). 12
13
(19) Taro-wa [[ringo-ga sara-no ue-ni ar-u]-no]-o mi-ta. Taro-Top apple-Nom plate-Gen on exist-Pres-NML-Acc see-Past 'There was an apple on the plate, and Taro saw it.' IHRC 'Taro saw the scene of an apple being on the plate.' Complement (20) a. Taro-wa [[aoi ringo-ga sara-no ue-ni ar-u]-no]-o tot-ta. Taro-Top green apple-Nom plate-Gen on exist-Pres-NML-Acc pick.up-Past 'There was a green apple on the plate and Taro picked it up.' IHRC b. Taro-wa [[ringo-ga sara-no ue-ni ar-u]-no]-o yokisitei-ta. Taro-Top apple-Nom plate-Gen on exist-Pres-NML-Acc anticipate-Past 'Taro anticipated that there was an apple on the plate.' Complement (21) a. [[Soba-no nagaisu-no ue-de Lucky-ga nete-i-ta]-no]-ga pro Nearby-Gen couch-Gen on Lucky-Nom sleep-Prog-Past-NML-Nom ookina akubi-o si-ta. big yawn-Acc do-Past ‘As Lucky was sleeping on a nearby couch, she made a big yawn.’ b. [[[[[Lucky-ga atama dake dasite-i-ru]-no]-no usiro]-ni Lucky-Nom head only show-Prog-Pres-NML-Gen behind-at ututtei-ru]-no]-wa kimi ka? appear-Pres-NML-Top you Q ‘Is it you in the photo standing behind [the scene in which Lucky is sticking out her head]?’ As a parting shot, we wish to offer a remark on (22a), which was brought up in Grosu (2012) in support of the thesis that maximalizing, i.e., singleton-denoting, relatives do not stack with proper intersective import. Thus observe that this example does not have the meaning of [ii], which implies that Mary wrote more than one interesting paper, that John read/improved one of them, and that the outcome of John's work was accepted by LI; rather, it implies that Mary wrote one interesting paper, and that that paper was read/improved by John and subsequently accepted by LI. The meaning in [i] is the one assigned to it by Grosu (2012), and it correctly captures the implication just noted, but it is not entirely correct for another reason, to which we turn directly. (22) a. [[John-ga [[Mary-ga omosiroi ronbun-o kaita]-no]-o John-Nom Mary-Nom interesting paper-Acc wrote-NML-Acc {yonda, kaizensita}]-no]-ga sokuzani LI-ni zyurisareta. {read, improved}-NML-Nom unhesitatingly LI-Loc was-accepted [i] ‘Mary wrote an interesting paper, John {read, improved} it, and it was unhesitatingly accepted by LI.’ ≠ NOT: [ii] 'The interesting paper that Mary wrote that John {read, improved} was unhesitatingly accepted by LI.' b. ?*[[[John-ga [[Mary-ga omosiroi ronbun-o kaita]-no]-o ni-hon John-Nom Mary-Nom interesting paper-Acc wrote-NML-Acc 2-Cl {yonda, kaizensita}]-no]-no kekkan]-ga sokuzani siteki-sare-ta. 13
14 {read, improved}-NML-Gen defect-Nom promptly point.out-Pass-Past ‘Mary wrote two interesting papers, John {read, improved} them, and their defects were promptly pointed out (by someone)’ Thus, one thing that Grosu (2012) failed to notice is that the acceptability of this example is puzzling if it consists of two IHRCs, one of which serves as the IH of the other, since, in view of the fact that both IHRCs are definite, the higher one seems to have a definite referential expression as its IH, contrary to what has been argued for in this section. The solution of the puzzle lies, we submit, in the fact that both structures assumed to be IHRCs could in principle be adverbials, so it is sufficient, e.g., to assume that the higher structure is adverbial for the acceptability of (22a) (interpreted as Mary wrote an interesting paper, and John having {read, improved} it, it was unhesitatingly accepted by LI) to be no longer puzzling. To show that the key to the puzzle lies in homophony with adverbials, we offer (22b). In this case, both structures are unambiguously IHRCs, the lower one, because it is a SHRC, and the higher one, because it bears Genitive Case. This example is infelicitous, and the puzzle is thus solved. In section 7, we will suggest a principled reason for the infelicity of referential IHs.
6. Are there 'change' IHRCs? This question arises in connection with data like (23), which were brought up by Hoshi (1995) and came to be called ‘change’ IHRCs in the subsequent literature. They are so called because the denotatum of the IHRC is characterized not by a syntactically represented nominal within the relative, but rather by the result of a process of change described by (some sentence within) the relative. (23)
John-wa [[Mary-ga (gozentyuu-ni) ringo-o sibottekure-ta]-no]-o John-Top Mary-Nom (in-the-morning) apple-Acc squeeze-Past-NML-Acc (gogo-ni) hitoikide nomihosi-ta. in-the-afternoon in-a-gulp drink.up-Past ‘Mary squeezed apples (in the morning), and John drank it [= the juice produced by squeezing the apples] in a gulp (in the afternoon).’
Now, the existence of change IHRCs was recently challenged by Ken Hiraiwa (p.c.) on the grounds that the constructions to which this status has been attributed are, in his view, invariably interpretable as 'gapless light-headed EHRCs', i.e., EHRCs headed by a 'light' element, in particular, the (semantically non-empty) pronominal no, which is compatible with inanimate entities or children, but is offensive when applied to venerable human beings, and can furthermore, unlike the (semantically vacuous) nominalizer no, be modified by adjectives (see Tsubomoto 1981 and Ohara 1994 for the observation that the nominalizer no in IHRCs is incompatible with adjectival modification, and see Hoshi 1995: 122, fn.5 for some discussion and a warning concerning the structural ambiguity between the genuine IHRCs and the gapless light-headed EHRCs in Japanese). We believe, however, that Hiraiwa's challenge can be successfully refuted, a task to which we
14
15 turn directly. We propose to argue that both change IHRCs and gapless light-headed EHRCs exist by showing that either of them can be unambiguously identified.2 Thus, we can be sure that a particular constituent is a gapless light-headed EHRC if it felicitously allows adjectival modification of no (as already noted), and also if it can be felicitously used as a topic, because incontrovertible IHRCs cannot (see Itô 1986, Kuroda 1999, Matsuda 2002 for the contrast between EHRCs and IHRCs). This is brought out by the facts in (24)-(25). Thus, the bracketed constituent in (24a), which bears Nominative Case, is either a nonchange IHRC or an adverbial, while the constituent in (24b), which bears a topic marker, can be neither. (24) a. [[Hon-ga san-satu kirei-ni narande i-ta]-no]-ga (pro) kyuuni otiteki-ta. book-Nom three-Cl neatly lining be-Past-NML-Nom suddenly fall-Past ‘Three books were lined up neatly, and they suddenly came down.’ IHRC ‘While three books were lined up neatly, they suddenly came down.’ Adverbial b.*[[Hon-ga san-satu kirei-ni narande i-ta]-no]-wa kyuuni otiteki-ta. book-Nom three-Cl neatly lining be-Past-NML-Top suddently fall-Past ‘(As for) the three books that were lined up neatly, they suddenly came down.’ (We note, for completeness, that, in contrast to -wa as topic marker, -wa as contrastive focus marker is compatible with IHRCs, as originally pointed out by Kuroda 1976-77, 1999, but it does not allow an adverbial construal). In contrast, the constituent in (25), which is unambiguously a gapless light-headed EHRC, due to the adjectival modification of no, happily allows a topic marker. (25) [[[Sally-ga orenzi-o sibotte-oi-ta] oisisoona]-no]-wa Sally-Nom orange-Acc squeeze-Aux-Past delicious-looking-one-Top John-ga itadai-ta. John-Nom have-Past ‘John had the delicious-looking orange juice such that Sally squeezed oranges (to obtain it).’ To show that change IHRCs also exist, we will exploit the fact (for which we will suggest an explanation in section 7), that an incontrovertible IHRC and an incontrovertible EHRC cannot be felicitously conjoined with a single Case marker borne by the coordination. Thus, a full EHRC and a gapless light-headed EHRC may conjoin, as in (26):
Tonosaki (1996) and Tonosaki (1998) use “change-HIRC” and “change-Relatives”, respectively. It is to be noted, however, that they in fact refer to (gapless) light-headed EHRCs in Japanese that happen to involve change-of-state-inducing predicates. Thus, they should be strictly differentiated from bona fide ‘change’ IHRCs discussed in this paper (see Hoshi 1995 on this point). For a recent analysis of (gapless) light-headed EHRCs, see Hiraiwa (2012). 2
15
16 (26) John-wa [[[Mary-ga teeburu-no ue-ni oite-oi-ta] ringo] to John-Top Mary-Nom table-Gen on put-Aux-Past apple and [[[Sally-ga orenzi-o sibotte-oi-ta] oisisoona]-no]]-o itadai-ta. Sally-Nom orange-Acc squeeze-Aux-Past delicious-looking-one-Acc have-Past ‘John had the apple that Mary put on the table and the delicious-looking orange juice such that Sally squeezed oranges (to obtain it).’ Two non-change IHRCs may also conjoin, as in (27) (= Hoshi 1995: 267): (27) Ken-wa [[[[Risa-ga teeburu-no ue-ni ringo-o oite-oi-ta]-no] to Ken-Top Risa-Nom table-Gen on apple-Acc put-Aux-Past-NML and [[[Erika-ga sara-no ue-ni mono-o oite-oi-ta]-no]]-o totte tabe-ta. Erika-Nom dish-Gen on peach-Acc put-Aux-Past-NML-Acc pick.up eat-Past ‘Risa put an apple on the table and Erika put a peach on the dish, and Ken picked up and ate them.’ But a non-change IHRC may conjoin neither with a full EHRC, nor with a gapless lightheaded EHRC, as shown in (28) and (29) respectively. (28) *Keesatu-wa [[[ [e] nigete-i-ta] otoko]] to [[[doroboo-ga san-nin police-Top run.away-Prog-Past man and thief-Nom three-Cl ie-no usiro-ni kakurete-i-ta]-no]]-o tukamaeta. house-Gen back-at hide-Prog-Past-NML-Acc caught ‘The police caught the man who was running away, and three thieves were hiding behind the house and the police caught them.’ (29) *Bill-wa [[[John-ga wain-o dasitekure-ta]-no] to Bill-Top John-Nom wine-Acc serve-Past-NML and [[[Mary-ga ringo-o sibottekure-ta] oisisoona]-no]]-o Mary-Nom apple-Acc squeeze-Past delicious-looking one-Acc zenbu non-da. all drink-Past ‘John served wine and Billi drank it all and (hei) drank all the delicious-looking apple juice such that Mary squeezed apples (to obtain it).’ With these facts in mind, consider now (30), in which an incontrovertible non-change IHRC is conjoined with a constituent which, in isolation, could be either a gapless lightheaded EHRC or a change IHRC. If the second conjunct could only be a gapless lightheaded EHRC, (30) should be infelicitous, like (29). If the second conjunct allows an interpretation as an IHRC, (30) allows a felicitous interpretation, like (27). In that case, the second conjunct is without doubt a genuine change IHRC. The facts support the latter: (30) is acceptable. Hence genuine change IHRCs do exist. (30) Bill-wa [[[John-ga wain-o dasitekure-ta]-no] to Bill-Top John-Nom wine-Acc serve-Past-NML and [[Mary-ga ringo-o sibottekure-ta]-no]]-o zenbu non-da. 16
17 Mary-Nom apple-Acc squeeze-Past-NML-Acc all drink-Past ‘John served wine and Mary squeezed apples, and Bill drank them (= the wine and apple juice) all.’
7. The Revised Relevancy Condition and its integration into a formal analysis 7.1. Five questions In sections 2-6, the issues (i)-(v) of section 1 were prominently addressed. In this section, we will address the issues (vi)-(vii) of that section, that is to say, the reasons for preferring a particular type of analysis over others, on grounds of empirical adequacy and explanatory power. We list below the properties of Japanese IHRCs that were established in earlier sections and that an adequate analysis needs to capture, as well as a number of queries that arise in connection with these properties and that a satisfactory analysis ought to shed some light on. (31) I. The IH needs to be an (existentially or otherwise) quantified nominal with relative-internal scope. It cannot be a (definite) referential expression, and in particular a proper name. Why? II. IHRCs are necessarily interpreted as definite descriptions. III. If the IH is existentially quantified, it may exhibit cancellable numerical implicatures, just like comparable expressions in discourse. Why? IV. The IH, whether syntactically represented or not (as in change IHRCs), may not be internal to an island, in particular, to a complex NP or adjunct island. V. How exactly does the Relevancy Condition constrain IHRCs, distinguishing them from discourses and EHRCs? And how can it be optimally integrated into a formal analysis? 7.2. Question III revised (31 III), which arises out of the parallelism between (14'b) and (15) pointed out in section 4, requires some modification. (32)-(33) show essentially what (14'b)-(15) do, namely, that existentially quantified nominals may exhibit cancellable numerical implicatures both when serving as E-type antecedents in discourse and when serving as internal heads in IHRCs. (32) Doroboo-ga huta-ri nige-teite Anthony-wa karera-o tukamae-ta. thief-Nom two-Cl run.away-Prog Anthony-Top they-Acc catch-Past Sikasi san-nin-me-no doroboo-mo nige-teite Anthony-wa kare-o but three-Cl-th-Gen thief-also run.away-Prog Anthony-Top he-Acc tukamae-ru koto-ga deki-nakat-ta. catch-non-Past thing-Nom be.able-Neg-Past 'Two thieves were running away, and Anthony caught them. But a third thief was also running away, and Anthony did not manage to catch him.' 17
18
(33) Anthony-wa [[doroboo-ga huta-ri nige-teiru]-no]-o tukamae-ta. Anthony-Top thief-Nom two-Cl run.away-Prog-NML-Acc catch-Past Sikasi san-nin-me-no doroboo-mo nige-teite Anthony-wa kare-o but three-Cl-th-Gen thief-also run.away-Prog Anthony-Top he-Acc tukamae-ru koto-ga deki-nakat-ta. catch-non-Past thing-Nom be.able-Neg-Past 'Two thieves were running away, and Anthony caught them. But a third thief was also running away, and Anthony did not manage to catch him.' (34) shows that a minimally different definite EHRC carries no comparably cancellable implicature, with the result that this example is felt to be self-contradictory. (34) Anthony-wa [[ [e] nige-teiru] huta-ri-no doroboo]-o tukamae-ta. Anthony-Top run.away-Prog two-Cl-Gen thief-Acc catch-Past #Sikasi san-nin-me-no doroboo-mo nige-teite Anthony-wa kare-o but three-Cl-th-Gen thief-also run.away-Prog Anthony-Top he-Acc tukamae-ru koto-ga deki-nakat-ta. catch-non-Past thing-Nom be.able-Neg-Past 'Anthony caught the two thieves that were running away. #But a third thief was also running away, and Anthony did not manage to catch him.' For completeness, we note that in Grosu (2010, fn. 5), it was stated that the EHRC in this example is necessarily definite, based on judgments provided by Akira Watanabe. In a recent p.c., Akira Watanabe indicates that the definite reading is only a strong preference for him, and that an indefinite reading may be made more salient by placing the numeral expression huta-ri-no in initial position within the relative clause, in which case there is no self-contradiction. The author of this paper whose name is listed second basically agrees with these intuitions, and – importantly – fully corroborates the intuition that the example is self-contradictory on the definite interpretation. Now, the fact that (34) exhibits no cancellable numerical implicature is not surprising, because such implicatures are normally restricted to existentially quantified expressions, and there is no such expression within the relative clause in (34). At the same time, the presence of such an implicature in (33) is at least prima facie surprising. To see this, note that both the IHRC in (33) and the EHRC in (34) are interpreted as definite descriptions (on the reading of the latter that is relevant here). In general, the definite article (whether overt or null) translates as the Sharvy-Link sigma, which differs from the sum operation in being presuppositional. That is to say, the definite article presupposes that the unique maximal entity obtained by summing up the members of its predicate is in this predicate. This means that the EHRC in (34) presupposes that the maximal sum of thieves running away had cardinality 2, hence, the infelicity of the second sentence, which contradicts this presupposition. Since the IHRC in (33) is also a definite expression, why can the assumption that there were exactly two thieves who were running away be cancelled?
18
19 The question can be sharpened by forming the EHRC in (35) on the basis of the first sentence in (15) (repeated below for convenience), and by comparing (33) with (35). (15) There were two books on the desk, and Bill took them to the library. But there was also a third book on the desk, and Bill did not take it to the library. (35) Bill took to the library [the two books that there were [e] on the desk]. #But there was also a third book on the desk, and Bill did not take it to the library. The bracketed definite DP in (35) is an entity-denoting 'amount' relative construction, whose denotation is constructed on the basis of an existentially quantified nominal (for analytical details, see Grosu & Landman 1998, ms., and references therein). In contrast to (33), whose meaning is also constructed on the basis of an existentially quantified nominal, the exactly-assumption has pre-suppositional status here. This state of affairs is at least prima facie puzzling, and we thus replace (31 III) with (31 III’): (31) III’ If the IH is existentially quantified, it may exhibit cancellable numerical implicatures, just like comparable expressions in discourse, and unlike prima facie comparable (definite) entity-denoting amount EHRCs in English. Why? We call the result of replacing III by III’ in (31) (31’). 7.3. The five questions in earlier analyses With this modification in mind, we now turn to a consideration of (31'). As far as we are aware, most of the earlier analyses of Japanese IHRCs have attempted to grapple with only proper subsets of the issues raised in (31'). For example, Watanabe (1992) focused exclusively on (31' IV), and proposed, in essence, that when some nominal expression functions as IH, a null operator that undergoes cyclic A-bar raising is launched from the Spec of its maximal extended projection. 3 This technique is inapplicable to change IHRCs, where there is no syntactically represented IH, and furthermore sheds no light on the fact that definite referential expressions are not felicitous IHs, since there seems to be no clear theoretical reasons for banning the occurrence of a null operator in their highest Spec, and it is thus in principle possible for the null operator to be launched from there. Crucially, the semantically understood and syntactically un-represented IH of a change IHRC may not be inferred from a sentence internal to a syntactic island, in particular, from an EHRC properly contained within the IHRC, as illustrated in (36). (36)*Mary-wa [[Bill-ga [[e] oisisoona gureepu-o sibotta] kokku-ni Mary-Top Bill-Nom delicious grapes-Acc squeeze-Past cook-Dat tyuui-o muke-sase-ta]-no]-no azi]-o totemo yorokonde tanosin-da. 3
See Itô (1986) and Ishii (1989) for alternative analyses that assume IH-raising after overt syntax. 19
20 attention-Acc turn-CausPast-NML-Gen taste-Acc with-great-pleasure enjoy-Past ‘Bill drew Mary’s attention to a/the cook who squeezed some delicious grapes, and she greatly enjoyed the taste of it [= the resulting juice].’ This island violation cannot be captured under Watanabe's approach, because there is no syntactic position for launching the null operator. Other earlier analyses focused on (31' I-II), and to some extent on (31' V), either ignoring (31' IV) or seeking to deny its reality. Such analyses were put forward, e.g., in Hoshi (1995), Shimoyama (1999, 2001), and Kim (2007), and their way of dealing with (31' I-II) was, essentially, to assume an E-type anaphoric relation between the IH (serving as antecedent) and a relative external (null) anaphor. Kim also attempted to integrate aspects of the Relevancy Condition into her formal analysis, with limited success, in our view (for a detailed critique, see Grosu 2010). This approach, unless augmented with further assumptions, fails to capture (31' IV), since discourse anaphoric dependencies are notoriously island-insensitive (at least, in all the languages we know of), and furthermore sheds no obvious light on the infelicity of definite referential IHs, since discourse anaphora incorporates the referential variety (but see Shimoyama 1999, 2001 for an attempt to account for such indefiniteness restrictions on IH under her particular E-type anaphora analysis). 7.4. The Relevancy Condition The only analysis we know of that attempts to address all the issues in (31') is the one initially put forward in Grosu & Landman (2012) and subsequently refined in Landman (ms.). Prior to outlining the gist of that analysis and evaluating its explanatory potential with respect to the queries raised in (31'), we will note the need for certain modifications in the original version of the Relevancy Condition (repeated below for convenience), some of which were noted by a number of earlier researchers (e.g., Kim 2002, Kim 2008, Grosu 2010). (6) a. The Relevancy Condition For an internally headed relative to be acceptable, it is necessary that it be interpreted pragmatically in such a way as to be directly relevant to the pragmatic content of its matrix clause. b. Sub-condition The two events represented by the internally headed relative and the matrix clause involve the same temporal interval and the same location. Concerning (6a), we propose to understand the notion 'directly relevant' in essentially the sense originally characterized and illustrated by Kuroda, that is, as the requirement that some relation or relations exist(s) between an event (or state; see below) described by the relative and an event described by its matrix such that the two events are easily construable, possibly with the help of pragmatic bridging, as a coherent, unified superevent; for example, if the event in the relative is construable as having been carried out 20
21 with the purpose of facilitating the implementation of the event in the matrix (for a variety of illustrations, see Kuroda 1975-76). For completeness, we note that there have been a number of subsequent attempts based on Japanese and/or Korean data to characterize the notion at issue, but we are not sure whether these attempts always targeted genuine IHRCs; for example, Kim (2008) proposes that direct relevance is satisfied only by a circumstantial, cause-effect, or concessive/contrastive relation, and she further proposes that this requirement is limited to a class of constructions she dubs 'event-central' IHRCs (the remaining constructions being dubbed 'entity-central'), but our impression, based on a close examination of her examples and of her description of their properties, is that both types of construction she proposes to identify are in principle open to both IHRC and adverbial construals; we are thus hesitant to draw any conclusions from her paper. Be this as it may, we will not delve any deeper into this matter here, and will content ourselves with Kuroda's original characterization. As for (6b), we wish to draw attention to a number of modifications and refinements that we view as necessary, and whose effect is to make the condition less restrictive. A first point is that the condition needs to be generalized to eventualities, because it may also be satisfied by states, as, e.g., in (37). (37) [[Kono sara-no ue-ni san-ko-no kookyuuna ringo-ga ar-u]-no]-ga this plate-Gen on three-Cl-Gen high-quality apple-Nom exist-Pres-NML-Nom oisisoo da. delicious-looking Cop ‘There are three high-quality apples on this plate, and they are delicious-looking.’
A second point is that the IH need not be a participant in an eventuality described by the entire relative clause, rather, it may also be a participant in an eventuality described by a clause properly contained within the relative, as in (38) (= (39a) in Watanabe 2003) (for further illustration, see, e.g., (29)-(30) in Grosu 2010). The Relevancy Condition must thus be modified to allow such data, in particular, by referring to the eventuality described by the sentence that immediately contains the IH, rather than to the eventuality described by the entire relative clause. (38) Mary-ga [[[John-ga [zibun-no gakusei-ga zyuuyoona kasetu-o Mary-Nom John-Nom self-Gen student-Nom important hypothesis-Acc teian-si-ta to] zimansite-ita]-no]-no kekkan]-o siteki-si-ta. propose-do-Past Comp boast-have-Past-NML-Gen defect-Acc point.out-do-Past 'John had boasted that his student proposed an important hypothesis and Mary pointed out a defect in it.' Constructions like (38) are also in need of some pragmatic bridging, which is not specific to the Relevancy Condition, since it also applies to minimally different discourses like the English translation of this example. Thus, note that while the IHRC is defined at the intensional indices of the matrix (i.e., those of the speaker's beliefs), the sentence that 21
22 immediately contains the IH is defined at the indices of what John boastfully stated, which, at least in principle, may or may not intersect with the indices of the matrix. Nonetheless, unless the speaker tacitly assumes that John's boast is true at least to the extent that his student proposed a hypothesis (albeit not necessarily an important one), (s)he cannot coherently assert that Mary pointed out a defect in it, because the presupposition of (unique) existence associated with the definite pronoun it would not be satisfied. A third extremely important point is that the eventuality within the relative clause that satisfies the sub-condition in (6b) (extended to worlds, as proposed above) need not be one described by an actual sentence included, properly or not, within the relative. That is to say, data like (2a) or (39), where the eventuality that satisfies the sub-condition is described by a sentence in which the IH is a verbal argument constitute only the simplest case. (39) Teki-wa [[syoogun-ga koohaisita ie-de zyuuyoona gokuhi kaigi-o enemy-Top general-Nom dilapidated house-at important secret meeting-Acc syoosyuusi-ta]-no]-o bakugekisi-ta. call-Past -NML-Acc bomb-Past ‘The general called an important secret meeting in a dilapidated house and the enemy bombed it (= the house).’ Thus, there exist a number of more complex types of situation, some of which we will now describe and illustrate. A first type of situation, prominently pointed out in Kim (2007) and subsequently further generalized in Grosu (2010), is found in constructions where a participant in an event described by a relative internal sentence is also found in a state resulting from that event, and it is the state, rather than the event itself, which satisfies the temporal and locational intersection requirement. Furthermore, as Landman (ms.) observes, there exist situations in which the event described by the matrix temporally precedes the event described by the relative without being temporally continuous with it, and in such cases, the temporal intersection requirement may be satisfied by a state resulting from the matrix event. Illustrations of these types of situation are provided in (38) and (40).4 4
An example in which the matrix event precedes the relative-internal event, and which exhibits a further interesting twist, was brought up in Grosu (2010); we reproduce his example (34) in (i) below. (i) Yamada-san-wa [[otonari-no musukosan-ga raigetsu kawaii onna-no-hito Yamada-hon.Top next.door-Gen son-Nom next.month pretty woman-Gen-person -to kekkon-suru]-no]-o tyoonai-no huzinkai-ni -with marriage-do.nonPast-NML-Acc neighbourhood-Gen women's club-Dat kanyuusiyootosi-ta. try.to.talk.into.joining-Past 'The next door neighbor's son is getting married to a pretty womani next month and Ms. Yamada tried to talk heri into joining the women's club in the neighborhood.' 22
23
(40) a. Taro-wa [[Yoko-ga asa kukkii-o Taro-Top Yoko-Nom morning cookie-Acc kat-ta]-no]-o ban-ni paatii-ni motteit-ta. buy-Past-NML-Acc evening-in party-to bring-Past "Yoko bought cookies in the morning and Taro brought them to the party in the evening." b. Yoko-wa [[Taro-ga ban-ni paatii-ni Yoko-Top Taro-Nom evening-in party-to kukkii-o motteit-ta]-no]-o asa katte-oi-ta. cookie-Acc bring-Past-NML-Acc morning buy-Aux-Past "Taro brought cookies to the party in the evening, and Yoko had bought them in the morning." In both (40a) and (40b), there is no temporal interval shared by the buying and the bringing events, but the buying event results in a state of the cookies that makes them available to Taro, and this state is construable as temporally continuous with the event of bringing them to the party (which, we may assume, is initiated by a sub-event of picking them up from the location where Yoko left them). The examples in (40) contrast in felicity with examples in which the state resulting from the earlier event does not last until the beginning of the later event (for an illustration, see (43a) and the preceding comment on it). As for (38), note that there is no need for Mary's demonstration of the defective status of the hypothesis to have immediately followed the proposal of this hypothesis; what ensures felicity in this case is that the hypothesis proposed by the student plausibly lingered after its proposal in awareness states in the minds of interested linguists, and plausibly a relevant such state temporally intersects with Mary's rebuttal (note, furthermore, that this example also arguably satisfies (6a) in that there is an obvious causal connection between the importance boastfully attributed by John to the hypothesis and Mary's puncturing of the balloon. What has just been said about (38) enables us to keep a promissory note we issued earlier in connection with (9) (repeated below for convenience). Note that (9) satisfies the Relevancy Condition for essentially the same reasons that (38) does, contrary to what Kitagawa (2005) suggested. If so, the deviance of (9) can only be attributed to a violation of the CNPC, as claimed in section 3. (9) *Mary-ga [[[John-ga [[ [e] atarasii kasetu-o teiansi-ta] gakusei]-o Mary-Nom John-Nom new hypothesis-Acc propose-Past student-Acc hidoku homete-ita]-no]-no akirakana kekkan]-o suguni siteki-si-ta. extravagantly praise-had-NML-Gen obvious defect-Acc promptly point.out-do-Past
Here, the state that temporally intersects with the future wedding event, the pretty woman being a participant in both of them, does not result from the event of Mrs. Yamada talking to her, but rather from an unexpressed, but contextually assumed, event of the neighbour's son designating her as his bride-to-be. 23
24 ‘John extravagantly praised the student [who had proposed a new hypothesis] and Mary promptly pointed out an obvious defect in it.' A second type of situation is provided by change IHRCs (discussed and illustrated in section 6), which differ from the previously described situation in that the participant shared by the relative and the matrix is found only in a state that results from an event described by an actual sentence, but not in the event itself. A third type of situation is illustrated by (41b), which may be compared with (41a). (41) a.#[[Taro-ga [DP2 [DP1 zibun1-no] tuma2]-no susi]-o Taro-Nom self-Gen wife-Gen sushi-Acc kyaku-ni dasi-ta]-no]-o kyaku-ga suguni home-ta. guest-Dat serve-Past-NML-Acc guest-Nom immediately praise-Past ‘Taro served to the guest the sushi of his wife and the guest praised her immediately after that.’ b. [[Taro-ga [DP2 [DP1 [daidokoro-no] zibun1-no] tuma2]-no Taro-Nom kitchen-gen] self-Gen wife-Gen susi]-o kyaku-ni dasi-ta]-no]-o kyaku-ga suguni sushi-Acc guest-Dat serve-Past-NML-Acc guest-Nom immediately yon-de home-ta. call-and praise-Past ‘Taro served to the guest the sushi of his wife, who was in the kitchen, and the guest called and praised her immediately after that.’ In both sub-cases of (41), the wife is neither a participant in the event of serving the sushi, nor does she play a role in any state resulting from this event. She is, admittedly, the Agent of the event of preparing the sushi, and this event undoubtedly provides a justification for the fact that the guest praised her, but cannot be viewed as satisfying the Relevancy Condition, because there is no temporal or locational intersection between this event and the sushi-serving event; hence, the infelicity of (41a). What distinguishes (41b) from (41a) is that in the former, but not in the latter, the wife is also present on location when she is called and praised by the guest. It is the state of her being in the kitchen that intersects temporally and locationally with the sushi-serving event, hence, the acceptability of (41b).5
5
The astute reader may be surprised by the acceptability of (41b), in view of what was said in section 5 about the infelicity of definite referential IHs. Shimoyama (2001, Chapter 3) notes that while data like (17b), whose IH is a proper name, are pretty bad, comparable data with a definite NP as IH, such as Isadora-no neko 'Isadora's cat', are more acceptable, and attributes this distinction to the fact that definite NPs, unlike proper names, are not necessarily referential, and may in principle also denote properties. In (41b), we may view zibun-no tuma as meaning something like “a wife of his own,” with the uniqueness of the wife being determined pragmatically. Thus, the same reasoning applies to cases that involve expressions such as Isadora-no neko ‘Isadora’s cat’ or kanozyo-no neko ‘her cat’, which can be analyzed as indefinites “a cat of Isadora’s” and 24
25 A fourth and last type of situation we will consider here is illustrated by (13). In this example, the IH, i.e., the book, is not a participant in the reading event whose existence is denied (if it were, it could not coherently function as IH; see discussion of (49b) below), but rather in an existing state of not having been read. The four types of situation we have just examined point to the conclusion that the IH need not be a participant (only) in the eventuality (more exactly, the set of eventualities) literally described by a relative-internal sentence, call it Eα (a term borrowed from Landman, this volume), and (only) in the eventuality described by the matrix, call it E (another term borrowed from Landman, this volume), but may also be a participant in (a set of) eventualities 'saliently associated' with Eα, in particular, resulting states, as in (40a) and change IHRCs, accompanying states, as in (41b), or states that exist without resulting from an event, as in (13), as well as in states accompanying or resulting from E, as in (40b); following Landman (ms.), we will call eventualities associated with Eα Eπ, and we will coin the term Eπ' for eventualities associated with E, noting that in 'simple' cases, such as (2b) and (39), Eα = Eπ and E = Eπ'. For convenience, we will use the symbol Erel to refer to the complex sub-eventuality that consists of Eα and Eπ, and the symbol Ematrix to refer to the complex sub-eventuality that consists of E and Eπ'. Using this terminology, we may now refine the Relevancy Condition as shown in (42). (42 I), which is inspired by the formulation in Landman (ms.), purports to capture the raison d'être of the Relevancy Condition, which, as we understand it, is to ensure that the relevant eventualities within the relative clause and its matrix in conjunction with their possible resulting and/or accompanying eventualities need to form a unitary and coherent super-eventuality; following Landman (ms.), the unified nature of the supereventuality is saliently brought out by characterizing it as a single group-eventuality, rather than as a plurality of eventualities. The function of (42 II) is to ensure that the Relevancy Condition is not only less restrictive than it was in its original formulation (in order to accommodate the four types of situation that were described above), but also more restrictive than the general desideratum for discourse coherence. Note that in addition to the requirement of a shared location and a shared temporal interval, which is taken over from (6b), we have added the requirement that the sub-eventualities that contain a shared participant must be defined in at least one shared world (this is the import of 'modal intersection'), and the need for this addendum will be demonstrated in section 7.5 (see examples (44)-(45) and discussion thereof). Conceivably, one may take the view that the three intersection requirements expressed in (42 II) are implicit in (42 I), but given the vagueness of the term 'natural super-eventuality', which seems no more precise than Kuroda's 'direct relevance', we prefer to state these requirements explicitly. (42) The Revised Relevancy Condition For an internally headed relative to be acceptable, it is necessary I. that Eα, Eπ, E and Eπ' be naturally interpretable as part of a single natural super-eventuality (a group eventuality), and II. that some (possibly proper) sub-part of Erel must intersect temporally, locationally, and modally with some (possibly proper) sub-part of Ematrix, “a cat of hers”, respectively, in Japanese. Hence, (41) need not be viewed as inconsistent with the ban on definite referential DPs as IHs that we proposed in section 5. 25
26 and the thus intersecting sub-eventualities must share a participant. 7.5. Putting relevancy to work The Revised Relevancy Condition accounts for a number of differences between IHRCs on the one hand and discourses and EHRCs on the other, to which we turn directly. Concerning (42 II), consider first the contrast between (43a) on the one hand and (43bc) on the other, which is attributable to the fact that temporal intersection is required for IHRCs, but not for discourses or EHRCs (note that neither the event of the cat coming in yesterday, nor the (temporary) state in which it found itself, intersects temporally with the event of the cat returning this morning). (43) a. ?*[[Haiiro-no neko-ga kinoo mado-kara haitteki-ta]-no]-ga gray-Gen cat-Nom yesterday window-from come.in-Past-NML-Nom kesa mata yatteki-ta. IHRC this-morning again come-Past 'A gray cat came in from the window yesterday, and it came back this morning.' b. Haiiro-no neko-ga kinoo mado-kara haitteki-ta. gray-Gen cat-Nom yesterday window-from come.in-Past Soitu-wa kesa mata yatteki-ta. Discourse that.fellow-Top this-morning again come.in-Past 'A gray cat came in from the window yesterday. It came back this morning.' c. [[ [e] kinoo mado-kara haitteki-ta] haiiro-no neko]-ga yesterday window-from come.in-Past gray-Gen cat-Nom kesa mata yatteki-ta. EHRC this-morning again come-Past 'The/a gray cat that came in from the window yesterday came back this morning.' Second, consider the contrasts in (44) (= (33) in Grosu & Landman 2012), in (45a-b) (= (53) and (52) respectively in Shimoyama (2001, Chapter 3), and in (45c-d) (= (6a-b) in Grosu 2010); as noted in Grosu (2010) and Grosu & Landman (2012), (45b) is ungrammatical for some speakers (for independent reasons, which are spelled out in Grosu 2010), but not for all, and we focus only on the judgments of consultants who accept (45b). (44) a. Hitorino insei-mo doyoobi-no paatii-ni ik-anakat-ta. no grad-student Saturday-Gen party-to go-Neg-Past Karera-wa zituwa uti-de taamu peepaa-o kaitei-ta. they-Top in-fact home-at term paper-Acc write-Prog-Past ‘No graduate student(s) came to the party on Saturday. They (i.e., the students were in fact writing term papers at home.’ b.*[[Hitorino insei-mo doyoobi-no paatii-ni ik-anakat-ta]-no]-ga no grad-student Saturday-Gen party-to go-Neg-Past-NML-Nom zituwa uti-de taamu peepaa-o kaitei-ta. 26
27 in-fact home-at term paper-Acc write-Prog-Past ‘*No graduate student(s) came to the party on Saturday, they (i.e., the nonexistent students at the party) were in fact writing term papers at home. (45) a. Honno suunin-no insee-sika doyoobi-no paatii-ni ik-anakat-ta. just a-few-Gen grad-student-sika Saturday-Gen party-to go-Neg-Past Karera-wa zituwa uti-de taamu peepaa-o kaitei-ta. they-Top in-fact home-at term paper-Acc write-Prog-Past 'Only a few graduate students came to the party on Saturday. In fact, they (= the other students) were writing term papers at home.' b.#[[Honno suunin-no insee-sika doyoobi-no paatii-ni just a-few-Gen grad-student-sika Saturday-Gen party-to ik-anakat-ta]-no]-ga zituwa uti-de taamu peepaa-o kaitei-ta. go-Neg-Past-NML-Nom in-fact home-at term paper-Acc write-Prog-Past '#Only a few graduate students came to the party on Saturday, and they (= those very students) were in fact writing term papers at home.' c. Tyoodo san-nin-no insei-ga doyoobi-no paatii-ni ki-ta. exactly three-Cl-Gen grad-students Saturday-Gen party-to come-Past Karera-wa zituwa uti-de taamu peepaa-o kaitei-ta. they-Top in-fact home-at term paper-Acc write-Prog-Past 'Exactly three graduate students came to the party on Saturday. In fact, they (= the other students) were writing term papers at home.' d. #[[Tyoodo san-nin-no insei-ga doyoobi-no paatii-ni ki-ta]-no]-ga exactly three-Cl-Gen grad-students Saturday-Gen party-to come-Past-NML-Nom zituwa uti-de taamu peepaa-o kaitei-ta. in-fact home-at term paper-Acc write-Prog-Past '#Exactly three graduate students came to the party on Saturday, and. they (= those very students) were in fact writing term papers at home.'
The discourses in (44a) and (45a,c) are salvaged by 'bridging' in something like the following way: when one states that no/few/only three students attended the party, it is plausible to assume that the speakers of these sentences expected most, possibly all, students to attend the party, and the anaphor karera 'they' may be construed as denoting those students who were supposed to come, but did not. This type of accommodation is not available with respect to (44b) and (45b,d); thus, (44b) has no coherent interpretation, and (45b,d) have only absurd interpretations which say that the same students were simultaneously at the party and at home. The uninterpretability of (44b) is addressed in Grosu & Landman (2012) (and in Landman, this volume), who propose that in the interpretation of Eα, the IH may not scope over negation. This means that the existence of a partying eventuality with students as participants is denied, so that there is no Eα such that a participant in it is also a participant in E. Note, however, that the non-existence of Eα does not exclude the existence of an Eπ in which the students who were at home attended the party at that time and which is defined only in worlds of expectation, but not in the real world. This is in fact exactly what salvages (44a): the antecedent of karera is defined in worlds of expectation, karera is defined in 27
28 the real world, and an anaphoric dependency exists between them. This salvaging strategy is, however, blocked in (44b), because it violates the modal intersection requirement of (42 II): Eπ and E, i.e., the sub-eventualities that share a participant, are only defined in disjoint (sets of) worlds. In contrast to (44b), (45b,d) may be construed as satisfying (42 II) in virtue of the fact that Eα and E are both defined in the real world and share a participant, in particular, few/exactly three students, but this interpretation is pragmatically absurd. In (44a,c), absurdity can be avoided by appealing to worlds of expectation, but this is not possible in (45b,d), for the same reason that it is not possible in (44b). Concerning (42 I), note that it enables E and/or Eπ' to restrict the interpretation of Eα and/or Eπ in IHRCs, but not in EHRCs, which are not subject to the Revised Relevancy Condition. This can be appreciated in relation to the contrast between (33) on the one hand and (34) with the EHRC construed as definite on the other. (33) Anthony-wa [[doroboo-ga huta-ri nige-teiru]-no]-o tukamae-ta. Anthony-Top thief-Nom two-Cl run.away-Prog-NML-Acc catch-Past Sikasi san-nin-me-no doroboo-mo nige-teite Anthony-wa kare-o but three-Cl-th-Gen thief-also run.away-Prog Anthony-Top he-Acc tukamae-ru koto-ga deki-nakat-ta. catch-non-Past thing-Nom be.able-Neg-Past 'Two thieves were running away, and Anthony caught them. But a third thief was also running away, and Anthony did not manage to catch him.' In (33), it is possible to construe the IH as denoting two thieves that were running away and that got caught by Anthony; the definite D that applies to the entire IHRC carries the presupposition that these were the only thieves who satisfied this entire characterization, and the continuation is felicitous, because it concerns a third thief who was merely running away, but did not get caught by Anthony. (34) Anthony-wa [[ [e] nige-teiru] huta-ri-no doroboo]-o tukamae-ta. Anthony-Top run.away-Prog two-Cl-Gen thief-Acc catch-Past #Sikasi san-nin-me-no doroboo-mo nige-teite Anthony-wa kare-o but three-Cl-th-Gen thief-also run.away-Prog Anthony-Top he-Acc tukamae-ru koto-ga deki-nakat-ta. catch-non-Past thing-Nom be.able-Neg-Past 'Anthony caught the two thieves that were running away. #But a third thief was also running away, and Anthony did not manage to catch him.' In contrast, the EHRC in (34) can only denote thieves that were running away, and if it is construed as definite, the definiteness operator presupposes that these were the only thieves who were running away, hence, the contradictory status of the continuation (a comparable account is applicable to (35)). What about (32)? (32) Doroboo-ga huta-ri nige-teite 28
Anthony-wa karera-o tukamae-ta.
29 thief-Nom two-Cl run.away-Prog Anthony-Top they-Acc catch-Past Sikasi san-nin-me-no doroboo-mo nige-teite Anthony-wa kare-o but three-Cl-th-Gen thief-also run.away-Prog Anthony-Top he-Acc tukamae-ru koto-ga deki-nakat-ta. catch-non-Past thing-Nom be.able-Neg-Past 'Two thieves were running away, and Anthony caught them. But a third thief was also running away, and Anthony did not manage to catch him.' Two juxtaposed sentences that form a discourse do not in general need to describe a unified super-event, but there is no ban on construing them in this way, either. In (32), where the two thieves are the antecedent of a definite anaphor, they are understood as being the only two thieves who were running away and were caught by Anthony. If there were additional thieves that satisfied this property and were not distinguishable from the two that were initially mentioned in any way, a subsequent reference to them is perceived as infelicitous (presumably because it violates the Gricean Maxim of Quantity). Similarly, an IHRC with comparable properties is infelicitous under comparable circumstances. This point is brought out by (46)-(47), which have the same initial sentences as (32)-(33), but whose continuations differ in the way just indicated. (46) Doroboo-ga huta-ri nige-teite Anthony-wa karera-o tukamae-ta. thief-Nom two-Cl run.away-Prog Anthony-Top they-Acc catch-Past #Sorekara kare-wa hoka-no nige-teiru doroboo-o tukamae-ta. Discourse then he-Top other-Gen run.away-Prog thief-Acc catch-Past ‘Two thieves were running away, and Anthony caught them. #After that, he caught the other running thieves.’ (47) Anthony-wa [[doroboo-ga huta-ri nige-teiru]-no]-o tukamae-ta. Anthony-Top thief-Nom two-Cl run.away-Prog-NML-Acc catch-Past #Sorekara kare-wa hoka-no nige-teiru doroboo-o tukamae-ta. IHRC then he-Top other-Gen run.away-Prog thief-Acc catch-Past 'Two thieves were running away, and Anthony caught them. #After that, he caught the other running thieves.’ In sum, in both IHRCs and comparable discourses, but not in definite EHRCs, it is possible to appeal to the context in order to add an implicit restriction to the IH/antecedent, and this restriction may give rise to a numerical implicature that is cancellable without violating the Gricean Maxim of Quantity. For the sake of completeness and in order to avoid possible confusion, we wish to stress that the presupposition of existence and uniqueness that came up in the two paragraphs that follow example (45) is a property of the (overt or null) definite article, and not of sentences or clauses. In (32), the initial sentence that translates as two thieves were running away has assertive, not pre-suppositional status, due to the fact that it forms a super-eventuality with the next sentence, which is itself asserted, so that a coherent unified super-eventuality can arise, we assume, only if the component sub-events have the same assertive/pre-suppositional status. Correlatively, we view the relative clause in (33) as asserted as well, because in view of the fact that the matrix is asserted, the relative clause needs to be construed as asserted as well if it is to form a coherent and unified 29
30 super-eventuality with the matrix, as required by the Revised Relevancy Condition (42). The assertive status of the relative clause of an IHRC was in fact pointed out in relation to comparable data from Korean by Chung & Kim (2003, section 3.5), who observed that IHRCs cannot be felicitously uttered if the information they provide is part of the prior context (see their (45) and their comments thereon). We can address at this point a fact that was noted in section 6, namely, that IHRCs do not happily conjoin with EHRCs. We suspect the reason lies in the fact that the former is subject to the Revised Relevancy Condition (42) and the latter is not. As a result, the relative clauses of IHRCs are necessarily part of a larger eventuality, while those of EHRCs are not. We conjecture that the coordination of these two types of objects cannot be assigned a coherent interpretation. The coherent super-eventuality requirement imposed by (42) can arguably also shed some light on two additional phenomena. One phenomenon, which was brought up earlier and is illustrated in (24), is that IHRCs are infelicitous when used as topics. We submit the reason is that topics need to be conceivable in isolation, i.e., independently of what is said about them, and this conflicts with the super-eventuality requirement. A second phenomenon, which was noted by Matsuda (2002), as well as by Chung & Kim (2003) in relation to comparable Korean data, is that IHRCs, in contrast to EHRCs, are not felicitous answers to certain types of constituent questions, in particular, to questions that call for an entity as answer, not for an eventuality. Thus, consider the facts in (48) and (49). (48) a. Anata-wa dare-o tukamae-ta no? you-Top who-Acc catch-Past Q ‘Whom did you catch?’ b. #(Watasi-wa) [[keemusyo-kara doroboo-ga san-nin nige-ta]-no]-o (I-Top) jail-from thief-Nom three-Cl run.away-Past-NML-Acc tukamae-ta no desu. catch-Past NML Cop ‘I caught the three thieves that ran away from jail.’ IHRC 'As three thieves ran away from jail, I caught them.' Adverbial c. (Watasi-wa) [[keemusyo-kara [e] nige-ta] doroboo-o san-nin] (I-Top) jail-from run.away-Past thief-Acc three-Cl tukamae-ta no desu. EHRC catch-Past NML Cop ‘I caught (the) three thieves that ran away from jail.’ (49) a. Keesatu-wa donoyoona zyookyoo-de doroboo-o tukamae-ta no. police-Top what-kind circumstance-under robber-Acc catch-Past Q ‘Under what kind of circumstance(s) did the police catch the robber?’ b. Keesatu-wa [[doroboo-ga okane-o nusun-dei-ta]-no]-o police-Top robber-Nom money-Acc steal-Prog-Past-NML-Acc tukamae-ta no desu. catch-Past NML Cop ‘The police caught the robber who was stealing money.’ IHRC 'As the robber was stealing money, the police caught him. Adverbial c. Keesatu-wa donoyoona zyookyoo-de doroboo-o huta-ri tukamae-ta no. 30
31 police-Top what-kind circumstance-under robber-Acc two-Cl catch-Past Q 'Under what kind of circumstance did the police catch the two robbers? d. Keesatu-wa [[doroboo-ga okane-o nusun-dei-ta]-no]-o huta-ri police-Top robber-Nom money-Acc steal-Prog-Past-NML-Acc two-Cl tukamae-ta no desu. catch-Past NML Cop ‘The police caught the two robbers who were stealing money.’ SHRC (48a) calls for an answer with an entity, and (48c) provides that entity, but (48b) does not, whether it receives an IHRC or an adverbial construal. Under an IHRC construal, (48b) asserts a coherent super-eventuality, and under an adverbial construal, it denotes two temporally and/or circumstantially related eventualities, neither of which constitutes an appropriate answer to (48a). In contrast, (49a) calls for an eventuality (i.e., the one that defines the circumstances under which the catching took place), and this eventuality is provided by the first part of (49b) under both the IHRC and the adverbial construal. For completeness, we provide the pair in (49c-d), where the answer includes a SHRC, whose felicity points to the conclusion that (49b) is felicitous not only under an adverbial, but also under an IHRC construal. 7.6. An integrated analysis Having established and justified a number of refinements of the original Relevancy Condition and having illustrated its explanatory potential with respect to a certain range of data, it remains to construct a formal analysis of IHRCs, into which the Revised Relevancy Condition is appropriately integrated. This task is addressed and carried out in detail in Landman (ms.), which refines the earlier analysis in Grosu & Landman (2012), and to which the reader is hereby referred. We confine ourselves in this section to sketching the principal ingredients of that analysis. The principal assumption of the analysis is that some sentential constituent within the relative is taken to denote Eα, and is provided with a null modifier, dubbed PPπ (the analytical 'heir' of Choose Role in Grosu 2010 and Grosu & Landman 2012), which carries out the following two tasks: (i) semantically, it identifies Eπ, selects a thematic role within it, and applies Quantificational Disclosure to the quantified object carrying this role by equating the role's value with a free variable, and (ii) syntactically, it launches a null operator that undergoes A-bar raising to the Spec of the relative clause, and which translates as an abstractor binding the variable created by disclosure; the predicate denoted by the relative CP undergoes Maximalization, a process found in a number of relative constructions (see Grosu & Landman 1998 and references therein), which coerces a definite construal of the null definite Determiner. For completeness, we note that in identifying Eπ, the null modifier appeals to the Revised Relevancy Condition, which involves reference to E, and thus may seem to require a violation of the principle of compositionality. Such a violation is avoided by assuming that the relevancy condition is introduced as a presuppositional condition on the null modifier involving a free event variable which gets identified with E at the level of the matrix (see Landman's ms. for details).
31
32 Importantly, the null modifier may not define Eπ as the denotation of a sentence properly contained within its sentential sister, even when this is semantically possible, that is to say, its search for an IH must be confined to a maximally local domain. This ensures that any violation of a sentential island is analytically detected. Note that launching the null operator in the way indicated avoids the problem that confronted Watanabe's (1992) analysis, which could not deal with island violation in change IHRCs. To be sure, the analysis does not explain why Japanese IHRCs are island-sensitive, since it is in principle possible to ensure abstraction over the variable created by disclosure without resorting to syntactic movement, and we do not even know whether IHRCs with the crucial properties of those found in Japanese are always island-sensitive, and in case this is so, whether an explanation exists (this is left as a topic for further research). Nonetheless, we wish to stress that there is nothing unnatural about assuming null operator movement under the proposed analysis, while there is arguably something unnatural about making such an assumption within an analysis that relies of E-type anaphora. For completeness, we note that the null modifier utilized in Landman (ms.) has PP status, and the null operator is launched from the null P's DP complement; in Grosu & Landman (2012), the counterpart of P was a functional head labeled Ch(oose) R(ole), and the null operator was launched from its Spec. As far as we can see, both mechanisms can do the intended job, and we express no preference for either of them, pending the discovery of strong empirical and theoretical motivation for a particular choice.6 We wish to conclude this section by noting that the mechanism of quantificational disclosure, which is an integral part of the analysis, sheds light on the fact that definite referential IHs are infelicitous (see section 5). In general, disclosure applies nonvacuously only to quantified DPs, since referential ones provide no justification for the procedure (see Dekker 1993). Independently of this fact, we note that the denotation of an IHRC is restricted not only by predicates internal to the IH (via equation), but also by the predicate defined by the remainder of the relative clause. Now, when the IH is referential, the restriction imposed by the latter predicate is vacuous, which is only possible if IHRCs may be appositive constructions. But appositive construals seem to be unavailable for IHRCs in general, and are in any event unavailable for Japanese IHRCs (pace Kuroda 1974, section 1.3, who suggested that IHRCs (in Japanese) may be semantically appositive). Thus, compare (50a) under an appositive construal of the EHRC with (51a), which contains a minimally different IHRC, and note the difference in truth conditions, brought out by the difference in felicity between the continuations in (50b) and (51b) respectively, which is traceable to the fact that the predicate nige-teiru 'running away' does not restrict the external head in (50a), but does restrict the internal head in (51a). (50) a. Sakuya 6
Anthony-wa [[ [e] nige-teiru] tyoodo huta-ri-no doroboo]-o
Another possibility is that the null operator in question is a topic-like element which seeks/chooses its “antecedent” as IH in an event-type clausal domain on the assumption that the null operator is based-generated at the Spec of TopP and undergoes A-bar movement to the Spec of ForceP in the CP-peripheral system in the sense of Rizzi (1997). We will leave investigation of this issue for another occasion. 32
33 last night Anthony-Top run.away-Prog exactly two-Cl-Gen thief-Acc tukamae-ta. Appositive EHRC catch-Past ‘Last night, Anthony caught exactly two thieves, who were running away. b.#Sosite sarani sakuya kare-wa rooya-no soto-de kakureteiru doroboo-mo and furthermore last night he-Top prison-Gen outside hide-Prog thief-also huta-ri tukamae-ta. two-Cl catch-Past ‘#And furthermore, last night, he also caught two thieves who were hiding outside the prison.’ (51) a. Sakuya Anthony-wa [[doroboo-ga tyoodo huta-ri last night Anthony-Top thief-Nom exactly two-Cl nige-teiru]-no]-o tukamae-ta. IHRC run.away-Prog-NML-Acc catch-Past 'Last night, exactly two thieves were running away, and Anthony caught them.' b. Sosite sarani sakuya kare-wa rooya-no soto-de and furthermore last night he-Top prison-Gen outside kakure-teiru doroboo-mo hutari tukamae-ta. hide-Prog thief-also two-Cl catch-Past 'And, furthermore, last night, he also caught two thieves who were hiding outside the prison. We stress that a comparable explanation is not available to formal analyses that rely on discourse anaphora, since the latter allows definite antecedents.
8. Conclusion We have provided in this paper novel empirical support for the following theses: [A] Genuine IHRCs exist, and also occur in a variety we called 'SHRCs.' [B] Change IHRCs exist, vindicating a mechanism with the essential properties of ChR, which crucially exploits the resources of the Revised Relevancy Condition. [C] IHRCs are invariably definite descriptions. [D] IHRCs exhibit island-sensitivity. [E] IHs of IHRCs cannot be definite referential expressions. We have also provided explanations for a number of prima facie puzzles that do not have obvious explanations under various alternative analytical approaches in the literature. We wish to conclude this paper by noting a topic of interest for subsequent research. The earlier literature has noted a variety of striking parallelisms between the IHRCs of Japanese and comparable constructions in Korean, with the result that some scholars have assumed that the two languages have IHRCs with the same properties (e.g., Kim 2007). However, the past literature on Korean IHRCs exhibits at least some of the confusion that has plagued some of the literature on Japanese IHRCs, and that we have drawn attention 33
34 to in this paper. It thus seems desirable to compare the two languages by taking as point of departure and standard for comparison the properties of Japanese IHRCs that were noted and argued for in this paper. A first step in this direction is taken in Grosu, Hoshi and Daeyoung (ms.). We are hopeful that additional colleagues will continue this project by further investigating the fine-grained properties of IHRCs in the two languages. Acknowledgments: We are extremely grateful to Fred Landman, whose detailed and penetrating remarks on an earlier version of this paper have led to significant improvements of both substance and presentation; to Akira Watanabe, for counterchecking the acceptability values of subtle judgments; and to Ken Hiraiwa, for drawing our attention to the prima facie challenge to the existence of change IHRCs posed by gapless light-headed relatives. We also wish to thank the audience at the workshop on internally-headed relatives held in November 2012 at ZAS, Berlin for a variety of helpful comments. None of these persons is in any way responsible for the use we have made of their ideas, and we assume sole and full responsibility for all the remaining faults in the paper.
References Chung, Chan and Jong-Bok Kim (2003) Differences between Externally and Internally Headed Relative Clause Constructions. Proceedings of the 9th International Conference on HPSG, 43-65. Culy, Christopher D. (1990) The Syntax and Semantics of Internally Headed Relative Clauses, Doctoral dissertation, Stanford University. Dekker, Paul (1993) Existential Disclosure. Linguistics and Philosophy 16, 561-587. Erlewine, Mitcho and Isaac Gould (2012) Domain readings of IHRCs, MIT talk. Grosu, Alexander (2010) The Status of the Internally-headed Relatives of Japanese/Korean within the Typology of Definite Relatives. Journal of East Asian Linguistics 19, 231-274. Grosu, Alexander (2012) Towards a More Articulated Typology of Internally Headed Relative Constructions: The Semantics Connection. Language and Linguistics Compass (to appear in issue No. 7) Grosu, Alexander, and Fred Landman (1998) Strange Relatives of the Third Kind. Natural Language Semantics 6, 125-170. Grosu, Alexander, and Fred Landman (2012) A Quantificational Disclosure Approach to Japanese and Korean Internally Headed Relatives. Journal of East Asian Linguistics 21, 159-196. Grosu, Alexander, Koji Hoshi and Daeyoung Sohn (ms.) The Japanese-Korean connection: A contrastive study of the inventory and properties of their IHRCs. Hastings, Rachel (2004) The Syntax and Semantics of Relativization and Quantification: The Case of Quechua. Doctoral dissertation, Cornell University. Hiraiwa, Ken (2012) The Mechanisms of Inverted Relativization in Japanese: A Silent Linker and Inversion. Journal of Linguistics 48, 345-388. Hoshi, Koji (1995) Structural and Interpretive Aspects of Head-internal and Headexternal Relative Clauses. Doctoral dissertation, University of Rochester. 34
35 Hoshi, Koji (1996) Multiple Case-checking in the Head-internal Relativization in Japanese. MIT Working Papers in Linguistics 29, 21-40. Ishii, Yasuo (1989) Head-internal Relative Clauses in Japanese. ESCOL ’89, 234-245. Itô, Junko (1986) Head-movement at LF and PF: The Syntax of Head-internal Relatives in Japanese. University of Massachusetts Occasional Papers in Linguistics 11, 109-138. Kadmon, Nirit. 1990. Uniqueness. Linguistics and Philosophy 13, 273-324. Kim, Min-Joo (2007) Formal Linking in Internally Headed Relatives. Natural Language Semantics 15, 279-315. Kim, Min-Joo (2008) Relevance of Grammar and Pragmatics to the Relevancy Condition. Language Research 44, 95-120. Kim, Yong-Beom (2002) Relevancy in Internally Headed Relative Clauses in Korean. Lingua 112, 541-559. Kitagawa, Chisato (2005) Typological Variation of Head-internal Relatives in Japanese. Lingua 115, 1243-1276. Kubota, Yusuke, and E. Allyn Smith (2007) The Japanese Internally Headed Relative Clause Is Not an E-type Pronoun. Formal Approaches to Japanese Linguistics 4, 149160. MITWPL #55, MIT Kuroda, S.-Y. (1974) Pivot-independent Relativization in Japanese I. Papers in Japanese Linguistics 3, 59-93. Kuroda, S.-Y. (1975-76) Pivot-independent Relativization in Japanese II. Papers in Japanese Linguistics 4, 85-96. Kuroda, S.-Y. (1976-77) Pivot-independent Relativization in Japanese III: Types of Japanese Relatives. Papers in Japanese Linguistics 5, 157-179. Kuroda, S.-Y. (1978) Case-marking, Canonical Sentence Patterns and Counter Equi in Japanese. In J. Hinds and I. Howard, eds., Problems in Japanese Syntax and Semantics, 30-51. Tokyo: Kaitakusha. Kuroda, S.-Y. (1999) Syubunaizai Kankeisetu. In S.-Y. Kuroda and M. Nakamura, eds., Gengo no Naizai to Gaizai, 27-103. Tokyo: Kurosio. Landman, Fred (ms.) Japanese Internally Headed Relatives. A hybrid analysis with Kuroda functions, Lombardian presuppositions and an internal scope mechanism. Matsuda, Yuki (2002) Event Sensitivity of Head-internal Relatives in Japanese. In N. Akatsuka, ed., Japanese/Korean Linguistics 10, 629-643. Stanford: CSLI. Mihara, Ken-ichi (1994) Iwayuru Syuyoobu Naizaigata Kankeistu-ni Tuite. Nihongogaku 13, 80-92. Murasugi, Keiko (1994) Head Internal Relative Clauses as Adjunct Pure Complex NPs. In S. Chiba et al., eds., Synchronic and Diachronic Approaches to Language: A Festschrift for Toshio Nakao on the Occasion of His Sixtieth Birthday, 425-437. Tokyo: Liber Press. Ohara, Kyoko Hirose (1994) A Correlative Analysis of So-called Internally-headed Relativization in Japanese. Paper presented at the Annual Meeting of LSA, Boston. Rizzi, Luigi (1997) The Fine Structure of the Left Periphery. In L. Haegeman (ed.), Elements of Grammar, 281-338. Dordrecht: Kluwer. Shimoyama, Junko (1999) Internally Headed Relative Clauses in Japanese and E-type Anaphora. Journal of East Asian Linguistics 8, 147-182.
35
36 Shimoyama, Junko (2001) Wh-constructions in Japanese. Doctoral dissertation, University of Massachusetts, Amherst. Tonosaki, Sumiko (1996) Change of State Head-internal Relative Clauses in Japanese. Studies in Language Sciences 2, 31-47. Graduate School of Language Sciences, Kanda University of International Studies, Chiba. Tonosaki, Sumiko (1998) Change-Relatives in Japanese. Journal of Japanese Linguistics 16, 143-160. Tsubomoto, Atsuro (1981) It’s All No: Unifying Function of No in Japanese. Proceedings of the Kansai Linguistic Society 11, 81-90. Watanabe, Akira (1992) Subjacency and S-structure Movement of Wh-in-situ. Journal of East Asian Linguistics 1, 255-291. Watanabe, Akira (2003) Wh and Operator Constructions in Japanese. Lingua 1143, 519-558.
36