<<

ANOTHER APPROACH TO BARE-NP ADVERBIALS AS NOMINALS

KEIICHIRO KOBAYASHI Kanto Gakuin University

Seemingly idiosyncratic properties of Bare-NP Adverbials (BNAs), in fact, follow from assumptions supported on independent grounds. Lex- ical Restrictions and Typological Restrictions follow from the assumption that derivations of BNAs start with structures with abstract nominal heads, and Semantic Extension and Semantic Inclusion from the assump- tion that BNAs are theta-marked by their own nominal heads. Besides, predictions about possible types of BNAs are made cross-linguistically, based on the former assumption, and the long-standing problem of "Case-assignment" of BNAs is resolved, based on the latter assumption.*

Keywords: inherent Case, feature elimination, null N, PP arguments, adverbials

1. Introduction In English, a group of phrases unaccompanied by prepositions appear as adverbials functioning as temporal modifiers, locative mod- ifiers, adverbials of direction and adverbials of manner, as shown in (1). (1) a. I saw her that day. (temporal) b. I saw her someplace warm. (locative)

* This paper is developed from Kobayashi (1999) and its preliminary version was also read at Yokohama National University. I would like to express my gratitude to Yoshio Endo and other participants there for comments and questions, which helped me a lot to clarify problems in my paper. My special thanks are due to Amado F. Garcia Ruiz and Maria Elena Salinas de Garcia for their patient discus- sion with me about crucial Spanish examples in this paper. I also thank Alan Bur- den for answering my questions about English data and Yutaka Hayashi for accom- modating me with easy access to English corpus. I am also grateful to two anony- mous EL reviewers for their insightful comments and suggestions on an earlier ver- sion of this paper. All remaining inadequacies are, of course, my own.

English 16: 2 (1999) 353-380 -353- (C) 1999 by the English Linguistic Society of Japan 354 ENGLISH LINGUISTICS, VOL. 16, NO. 2 (1999)

c. I was headed that way. (direction) d. I did it that way. (manner) Bare-NP adverbials (BNAs, hereafter) are not very productive expres- sions because the heads of the NPs are restricted to a small group of . Among the four types of BNAs, only temporal BNAs allow as their nominal heads a variety of common nouns, i.e. nouns referring to calendrical units. (2) John arrived that minute/hour/day/week/month/year/time. However, as shown in (3a), nouns like war, childhood, etc. cannot appear as nominal heads though they also refer to some types of time. The nominal heads of the remaining three types of BNAs are restricted almost to a single noun, i.e. place for locative BNAs, way for BNAs of direction and BNAs of manner, as shown in (1b-c) and (3b-d), respec- tively. Let us call this property Lexical Restrictions.1 (3) a. *I saw her that wart my childhoodlthat occasion. (temporal) b. *I saw her some location/some area/some venue. (locative) c. *I was headed that orientation/that course. (direction) d. *I did it that manner/that method/that fashion. (manner) Furthermore, the appropriate choice of nominal heads is not suf- ficient for the generation of BNAs. Examples in (4) show that the heads must be accompanied by modifiers or which are more contentful than an indefinite a (and its corresponding null counterpart for a [+plural] head) and a definite article the. We call this property Semantic Extension.2 (4) a. *I saw her a day/days/the day. b. I see her that day. (=(1a)) c. I saw her every day. d. I stayed there several/many/a few days. e. I will see her someday.

1 Larson (1985: 597) remarks that the noun direction can occur as a nominal head of BNAs of direction in his dialect. However, since my informants either reject it or prefer way to direction, we regard way as a unique nominal head of BNAs of direction. Cf. McCawley (1986) and Kobayashi (1987) for exceptions to Lexical Restrictions. 2 This point is due to Nagai (1986, 1987). ANOTHER APPROACH TO BARE-NP ADVERBIALS AS NOMINALS 355

2. Previous Analyses Bresnan and Grimshaw (1978), McCawley (1986) and Emonds (1987) regard BNAs as PPs of which the head P is either deleted or empty. Though the mechanisms to license null Ps differ among the analyses, BNAs are licensed in the same way as PP adverbials with overt pre- positions. For example, as the NP that day is Case-licensed and theta- marked by the preposition on in the structure of PP on that day, BNA that day in (1a) is licensed by an empty preposition. In this analysis, Lexical Restrictions are regarded as selectional restrictions imposed by the empty P on its complement NP. A problem of this analysis is that BNAs do not necessarily have cor- responding PP counterparts, as illustrated in (5). (5) a. We met him (*in) this morning. b. We met him (*in) every morning. Larson (1985) claims that BNAs are NPs appearing as adverbials in virtue of an exceptional Case property of their nominal heads; nominal heads of BNAs allegedly have Case feature [+F], which is percolated to the dominating NP and the NP [+F] can assign itself an Oblique Case. BNAs in (1) are properly Case-marked even in the absence of the pre- ceding prepositions because the head nouns day, place, way (Direction) and way (Manner) are specified as [+F]. On the other hand, since head nouns in (3) such as war, location, orientation and manner are not spec- ified as [+F], NPs they head remain Caseless in violation of "Case Filter." An empirical problem of this analysis pointed out by Emonds (1987) and Kobayashi (1987) is over-generation of examples in (6). (6) a. *It was forgotten that day. b. *It did not seem the time ripe. Here, NPs with [+F] heads are generated as objects in passive con- structions. The problem is that since NPs that day and the time can be Case-marked through [+F] feature of the head nouns day and time, they don't have Case-theoretic reasons for which to move to the position occupied by it and they remain in situ. To avoid this prob- lem, we have to stipulate that the activation of [+F] is suppressed in A- positions or that NPs having Oblique Case in virtue of [+F] heads re- ceive adverbial theta-roles obligatorily. Conceptually, the distinction of DPs in terms of Case feature [+F] is not tenable in the current framework of Minimalism (cf. Chomsky (1995)), where all DPs are introduced into derivations along with Case 356 ENGLISH LINGUISTICS, VOL. 16, NO. 2 (1999) features. If Case feature of BNAs were to be checked by the same feature of some head, a plausible candidate for the head would be, again, empty P. Although Stroik (1992) also claims that BNAs are generated as NPs, he argues that temporal BNAs and BNAs of manner are predicates. However, this claim contradicts with the fact in (4) that while NPs in the forms of a N, the N and Ns do not make BNAs, those in the forms of, say, several Ns, many Ns and a few Ns do. This fact runs counter to the general fact pointed out by Rothstein (1983: 102-105) that while a N, the N and Ns are allowed as predicates, as shown in (7a-c), sever- al Ns, many Ns and a few Ns are not, as shown in (7d, e). (7) a. We consider him a good doctor. b. We consider them good doctors. c. We consider John the best doctor. d. *We consider the boys several/many idiots. e. *We consider the men a few soldiers. Let alone Semantic Extension in (4), previous analyses have not satisfactorily explained Lexical Restrictions, because the selection of appropriate nominal heads either by empty Ps or by the specification [+F] is not much different from a mere stipulation of the facts in (1)-(3). Besides, no analysis has answered why BNAs are restricted to the four types in (1), i.e. why there are no BNAs of, say, accompani- ment, benefactive, source, etc. Let us call this fact Typological Re- strictions, which should not be confused with Lexical Restrictions. Nagai (1986, 1987) points out that the prepositions whose can be subsumed by BNAs are restricted. It is not the case, for exam- ple, that temporal BNAs freely include the semantics of any preposi- tions denoting time; while BNAs that time, that day, that month or three days mean at that time, on that day, in that month and for three days, respectively, they never mean before that time, since that day, af- ter that month, etc. Similar observations apply to other types of BNAs; someplace warm means in someplace warm but not, say, near someplace warm. Let us call this fact Semantic Inclusion.

3. Proposal In Sections 3.1 and 3.2, we introduce two assumptions to be adopted in our analysis, and claim that BNAs are generated as DPs as a result of their interaction in Section 3.3. ANOTHER APPROACH TO BARE-NP ADVERBIALS AS NOMINALS 357

3.1. PP in A-Positions In English PPs appear in subject positions, as shown in (8). (8) a. Between six and seven suits her fine. b. Across the road was swarming with bees. (Jaworska (1986: 355)) While Chametzky (1985), Jaworska (1986) and Neelman (1997: 107) claim that the subjects in (8) are PPs, Conway (1997) (following the claim in several of Joan Bresnan's papers) claims that they are NPs (which we regard as DPs), in which a null N takes PP as its comple- ment. Within the subject in (8a), null temporal N takes a temporal PP as its complement, and within the subject in (8b), null locative N takes a locative PP as its complement, as shown in (9). (9) a. [DP[NP 0 [ppbetween six and seven]]] b. [DP[NP 0 [pp across the road]]] These two different approaches to "PP-subjects" apply to the analyses of PP-arguments in general. In (10), PP-arguments occur as objects of , and in (11), they occur as objects of prepositions. (10) a. The campaigners planned until Christmas in detail. b. The new tenants are reclaiming behind the garage. (Jaworska (1986: 356)) (11) a. They waited until after midnight. b. The baby crawled from under the table. (ibid.) Assuming that all null elements are affixes, Conway suggests, as a pos- sibility of their licensing, that null Ns are licensed as affixes as a result of the embedded Ps' incorporating with them. With this covert head- movement applied in the structures of subjects in (9), the structures in (12) are derived, where the null N satisfies its morphological require- ment by attracting a licensing host. (12) a. [DP[NP [Nbetweeni [N0]] [ppti six and seven]]] b. [DP[NP [Nacrossi [N0] ] [ppti the road]]] We adopt Conway's analysis (or its corresponding feature-movement counterpart) of "PP-arguments." In Section 3.3, we will propose an analysis of BNAs, in which null Ns in "PP-arguments" are identified with the sources of nominal heads of BNAs.

3.2. Elimination of Inherent Case Takano (1996) and Fukui and Takano (1998) propose Case systems, in which derivations containing Case features converge at LF not only by eliminating them through checking, but also by rendering them in- 358 ENGLISH LINGUISTICS, VOL. 16, NO. 2 (1999) terpretable either at LF or PF. That is, according to Takano (1996) and Fukui and Takano (1998), if Case feature is rendered interpretable either at LF or PF, it need not be eliminated through checking. Spe- cifically, if Takahashi's (1993) claim (based on a suggestion made by Mamoru Saito) is adopted that the Case feature of Japanese verbs is in- herent because it is always linked to the thematic property "assign Theme/Patient," it is rendered interpretable at LF.3 On the other hand, Case feature of objects of Japanese verbs need not be eliminated through checking, either. They claim that Case particles attached to the objects are morphological realizations of Case features and the for- mer render the latter visible to Spell-Out and interpretable at PF. Hence, Case feature of the objects is removed from the derivation by Spell-Out and does not cause crashing at LF, either. In fact, in Japanese, Accusative Case feature of verbs and that of ob- jects of verbs must be rendered LF-interpretable and PF-interpretable, respectively, in the above-mentioned manner. Since Takano (1996) and Fukui and Takano (1998) assume (i) that the Case feature [assign Accusative Case] belongs to V, rather than the functional head v, (ii) that checking always takes place within the domain of a functional head, (iii) and that verbs do not raise in Japanese, the elimination of Accusative Case features by checking is regarded as impossible in Japanese. Though we adopt this mode of eliminating inherent Case features, there remains a question as to why Case feature of the objects of Japanese verbs cannot be rendered interpretable at LF; it should also be inherent because it is always linked to the thematic property "re- ceive Theme/Patient." We assume that if some Case feature of theta- assigners is inherent, the same feature of theta-assignees is also inher- ent. Accordingly, we propose the condition (13) to account for the fact that only the Accusative Case feature of verbs, and not of their ob-

3 The assumption that Accusative Case is inherent in Japanese leads Takahashi (1993) to deny the presence of ECM in Japanese. He argues that NP Mary-o in the apparent ECM construction in (i) is an object of the matrix , as shown in the structure in (ii). (i) John-ga Mary-o kasikoi to omotteiru. John-Nom Mary-Acc is smart that thinks (ii) John-ga Maryi-o [CP[IP proi kasikoi] to] omotteiru ANOTHER APPROACH TO BARE-NP ADVERBIALS AS NOMINALS 359 jects, is rendered LF-interpretable in Japanese. (13) The inherent Case feature of a head X is rendered LF-inter- pretable, iff it contains another head Y which assigns a theta-role associated with the Case feature. (13) accounts for the above question, since the D heads of the objects of verbs do not contain another heads assigning theta-roles associated with Accusative Case. On the other hand, for the Case feature of Japanese verbs to be ren- dered LF-interpretable under (13), we need following modifications to the above analysis. First, contra Takano (1996) and Fukui and Taka- no (1998), who claim that Case feature of Japanese verbs is located in the "main verb" V, we follow Collins (1997) in assuming that it is lo- cated in the "" v. Secondly, contra Takano (1996) and Fukui and Takano (1998), who claim that the main verb V in Japanese re- mains in situ, and contra Chomsky (1995) and Collins (1997), who claim that of V to v always takes place overtly, we assume that the [-V] feature of the light verb v attracts the main verb V in Japanese at LF.4 Since it is generally assumed that checking always takes place within the domain of a functional head, Takano (1996) and Fukui and Takano (1998) claim that Accusative Case feature of V in Japanese, which does not move to the functional head v, must be elim- inated in the above-mentioned fashion. In contrast, since we adopt their mode of Case-elimination and yet admit covert V-to-v movement, there must be an independent reason why the objects of Japanese verbs do not undergo checking. In this respect, we simply assume that for languages having Case-affixes, the Case-elimination through Spell-Out is more economical than that through checking, since the former does not need the movement of features. On the other hand, Case feature of Japanese verbs, which we assume is in v, is always rendered LF- interpretable for free, as a result of v' attracting V at LF, with (13) holding in a newly formed complex v.

4 Both of our claims in the text, i.e. the claim that Case features of Japanese verbs are located in the "light verb" v and the claim that raising of V to v occurs covertly in Japanese, are assumptions in this paper. However, as we will show shortly, we can account for various facts of BNAs, if we adopt these assumptions. We take this result to constitute an for the assumptions. 360 ENGLISH LINGUISTICS, VOL. 16, NO. 2 (1999)

3.3. Self-Theta-Marking If, as suggested by Takano (1996) and Fukui and Takano (1998), their mode of Case-elimination extends to inherent Cases in general, inherent Case features of any head that has attracted a relevant theta- marking head are rendered LF-interpretable in our analysis, regardless of whether the Case-features are actually eliminated through checking. Suppose, as in Takano (1996), that prepositional phrases are headed by a functional category p which takes PP as its complement. As we assume that the Oblique Case features of both p and the object of P in English are inherent because of their link with "assign adverbial theta- roles" and "receive adverbial theta-roles," respectively, the Case fea- ture of p is rendered LF-interpretable under (13), as a result of its attracting P, with Ps being relevant theta-assigners (cf. Section 4.4 for exceptional Ps). However, since English has no Case particle to ren- der the Case feature of the object of P interpretable at PF, Case fea- tures of both p and the object of P must be eliminated through check- ing. If this is the reason why English pP projection adopts checking system for the elimination of Case features, we predict that if an adver- bial DP contains both Oblique Case and theta-marking property and the condition (13) is satisfied within the DP, the Case feature of D, being inherent, can be eliminated even in the absence of p. Before considering such cases, let us propose our licensing condition of adver- bials. According to Larson (1985), phrasal categories are licensed as adver- bials by his Adverbial Theta-Role Assignment, which optionally assigns adverbial theta-roles to any phrasal category. Here, we propose a licensing mechanism (14), which differs radically from Larson's in that it licenses phrases as adverbials simply because their heads are theta- marking heads or because their heads contain theta-marking heads. That is, while adverbials are licensed as such from without in Larson (1985), they are licensed as such from within in our analysis.5 (14) XP can be licensed as an adverbial of a particular type iff the head X or another head Y which X contains assigns that type of adverbial theta-role.

5 The licensingcondition (14) is concerned only with the eligibility of XPs as adverbials; more devices are needed to account for the distributionof adverbials. ANOTHER APPROACH TO BARE-NP ADVERBIALS AS NOMINALS 361

(14) follows from our assumption that the requirement of Theta- Criterion differs in terms of the kind of theta-roles to be assigned; as far as argument-roles are concerned, it is a requirement for a theta- assignee to receive a theta-role, as suggested by Chomsky (1995: 347): as far as adverbial theta-roles are concerned, it is a requirement for a theta-assigning head to assign a theta-role. Accordingly, it is predicted that while heads can fail to assign argument-roles, as in the case of ex- ternal theta-roles in nominalizations, they never fail to assign adverbial theta-roles. Therefore, the assigners of adverbial theta-roles obliga- torily assign their theta-roles to the recipients. Suppose that theta- marking must take place within the maximal extended projection of theta-assigning heads. Then, if the heads which assign adverbial theta- roles fail to take complements, they can theta-mark their own maximal extended projections. In fact, self-theta-marking in this case is what is forced by Theta-Criterion imposed on theta-assigners. We don't have to introduce devices like feature-percolation, because as is supposed above, the structural condition on theta-marking is only loose enough to admit self-theta-marking. The reason why self-theta-marking does not take place in the case of argument-roles is that Theta-Criterion is not imposed on theta-assigners in such cases.6,7

6 It is not clear from the description in Chomsky (1995: Ch. 4) whether the der- ivation of examples like (ia), in which a theta-assigning head fails to assign its theta- role, crashes at LF, violating FI, or (ia) is simply gibberish. (i) a. *John put the book. b. John put the book on the desk. 7 Since Larson's (1985) Adverbial Theta-Role Assignment and our alternative (14) are licensing conditions on optional elements, it is quite natural that their ap- plication is optional. In this respect, one might wonder why Theta-Criterion on adverbial theta-roles requires theta-assigning heads to assign adverbial theta-roles obligatorily. It seems that this situation is not problematic if licensing conditions on adverbials and Theta-criterion for their heads should have independent status. However, we can speculate that the optionality of adverbials follows from our claim that heads of adverbials obligatorily assign theta-roles. If a head obligatorily assigns its theta-roles, its non-occurrence saves the derivation, resulting in the optionality of adverbials. On the other hand, non-occurrence of an argument does not save the derivation, because if a head assigning an argument theta-role appears without an argument, it results either in a nonconvergent derivation, violating FI, or in gibberish. In such an approach, we may be able to dispense with licensing con- ditions like (14). 362 ENGLISH LINGUISTICS, VOL. 16, NO. 2 (1999)

As for the membership of self-theta-marking nouns, we identify them with the null Ns shown in Section 3.1. More strictly, we assume that if a language has in its lexicon a certain type of non-overt noun which is an assigner of a particular adverbial theta-role, it is allowed to take one of the two possible forms. As one option, it is introduced into deriva- tions as a null N. We have already observed such cases in Section 3.1. Alternatively, the non-overt noun can take the form of an abstract noun. The crucial difference between a null N and an abstract noun is that while the former is an LF-affix and remains non-overt throughout derivations, the latter is not an affix and can be morphologically real- ized as an overt lexical item at PF. Since English has a non-overt temporal noun and a non-overt locative noun, they can occur as null N heads of "PP-arguments," as shown in 3.1. On the other hand, those non-overt nouns can be introduced into derivations as abstract nouns TIME and PLACE, respectively, as shown in (15a, b). (15) a. [DPthat [NPTIME]] b. [DPsome [NPPLACE]] The abstract nouns in (15) do not have to take PP-complements be- cause (i) being not LF-affixes, they don't have to attract prepositions at LF, and (ii) a theta-assigner can assign an adverbial theta-role to the maximal extended projection, of which it is the lexical head. Specifi- cally, TIME and PLACE theta-mark the whole DPs in (15). Under the condition to be considered in Section 4.1, the abstract nouns in (15) are spelled out morphologically as overt common nouns such as day and place. There won't occur Case-theoretic problems, either. Suppose that Ds in (15) happen to have Oblique Case. According to Longobardi (1994), N moves to D at LF, so that the structures (16a, b) are derived from (15a, b), respectively. (16) a. [DPTIME+that [NPt]] b. [DPPLACE+some [NPt]] In the newly formed complex Ds, Ds with (inherent) Oblique Case contain abstract nouns with a property of assigning adverbial theta- roles. Since the latter theta-roles happen to be associated with Oblique Case feature of Ds, condition (13) is met and the inherent Case features of Ds in (16) are rendered LF-interpretable. This situa- tion resembles that for Japanese verbs, except that while Case feature of Japanese verbs is that of assigners', Case feature of adverbial DPs seems to be that of recipients'. However, our condition of Case-elim- ANOTHER APPROACH TO BARE-NP ADVERBIALS AS NOMINALS 363

ination (13) is insensitive to this difference. Being licensed by (14), DPs in (16) corresponding to BNAs in (1a, b) are licensed as such at LF.8 Although we have to wait until Section 4.2, to show that English also has non-overt nouns corresponding to the abstract nouns DIRECTION and MANNER, the same analysis applies to BNAs of direction and manner in (1c, d). That is, the abstract nominal heads of DPs in (17) are spelled out as way(direction) and way(manner), respectively, and those abstract nouns adjoin to D at LF, as shown in (18). (17) a. [DPthat [NPDIRECTION]] b. [DPthat [NPMANNER]] (18) a. [DPDIRECTION+that [NPt]] b. [DPMANNER+that [NPt]] We will account for Lexical Restrictions by considering what kind of principle is at work to the morphological realization of abstract nouns, in Section 4.1.9

4. Accounts of the Facts In this section, we apply our analysis proposed in Section 3.3 to the facts of BNAs.

4.1. Lexical Restrictions The fact in (3) that NPs with inherently relevant adverbial meaning

8 Since Ds can be introduced into derivationswith any Case features, Ds (with null Ns and PP-complements)in the structure (9a, b) can have any structural Case feature that suits the surface positions of their dominating DPs. On the other hand, when the non-overt nouns are introduced into derivationsin the form of self- theta-markingabstract nouns, as in (15a,b), "governing"Ds virtually must have an inherent Case feature, i.e. Oblique Case feature. If the Ds have structural Case features, they cannot be rendered LF-interpretable through N-to-D movement. The dominating DPs in such cases cannot occur as arguments, either, since self- theta-markingDPs occurringas arguments violate Theta-Criterion. 9 As we mentioned in Section 2, the examples in (6) are over-generated due to the self-Case-marking property of the italicized DPs in Larson's (1985) analysis. In our analysis, the Case-features of the italicized DPs are eliminated only when the DPs are theta-marked by their own nominal heads. On the other hand, the itali- cized DPs are already theta-marked in their positionsby forgotten in (6a) and ripe in (6b), respectively. Therefore, those examples are correctly excluded by Theta- Criterion. 364 ENGLISH LINGUISTICS, VOL. 16, NO. 2 (1999)

can fail to occur as BNAs led Larson (1985:596) and Emonds (1987: 621)) to assume that the membership in the class of head nouns of BNAs cannot be reduced to semantic considerations. On the other hand, however, some semantic principle seems to be at work to allow all the nouns referring to calendrical units to appear as heads of tem- poral BNAs (cf. (2)). Naturally, the interpretations of null Ns are severely restricted. Ex- amples in (19) show possible interpretations of null Ns in the structures of (9). (19) a. [The [time/*war/*childhood [between six and seven]]] suits her fine. b. [The [place/?location/?area/?venue/*school [across the road]]] was swarming with bees. The facts in (19) lead us to the generalization (20). (20) Null Ns are interpreted as semantically the most neutral and the least contentful noun representing the adverbial theta- roles which they assign. (20) excludes nouns like war and childhood in (19a) as possible inter- pretations for null temporal N in (9a), because, in contrast with time, they denote more than a mere period of time; they denote a type of time, as is obvious from the fact that they can be paraphrased into such expressions as "the time when..." Similarly, the noun place is chosen as the least contentful noun replacing null locative N in (9b), because all the other nouns in (19b) denote more than just a place; location de- notes some place for some purpose, and area is a place delineated in the light of some of its characteristics. Since we assumed in Section 3.3, that null N and its corresponding abstract noun are different forms of the same non-overt noun in the lexicon, we expect that morphological realizations of abstract nouns should reflect the semantic interpretations of null Ns in (20). Hence, the generalization (21) should hold. (21) Morphological realizations of abstract head nouns of BNAs must reflect the semantic interpretations of the correspond- ing null Ns. Since the interpretations in (19) suggest that the noun time and place are the least contentful nouns to realize the abstract nouns TIME and PLACE, the latter are realized as the former, in accordance with the generalization (21). This claim is also supported by the facts in (22)-(23), though the judgment is subtle. ANOTHER APPROACH TO BARE-NP ADVERBIALS AS NOMINALS 365

(22) a. That time was a war/a childhood. b. ?That war/?That childhood was time. (23) a. That place is a location/an area/a venue/a school. b. ?That location/?That area/?That venue/?That school is a place. Since for an NP to be predicated of another NP, the former must be more specific in meaning than the latter, the fact that (22b) and (23b) are less acceptable than (22a) and (23a), respectively, suggests that the nouns time and place are the least contentful nouns denoting time and place among all the nouns in (22)-(23). On the same ground, the noun way is selected as the most appropri- ate noun to realize the abstract noun DIRECTION (cf. (ic) and (3c)), since it is less contentful than other nouns representing the same theta- role: nouns orientation and course suggest the presence of hidden agents more strongly than way. Cf. also (24). (24) a. This way is an orientation/a course. b. ?This orientation/?This course is a way. Way is also selected as the noun to realize the abstract noun MANNER (cf. (id) and (3d)); nouns like manner, method and fashion seem to suggest more strongly the presence of hidden agents than way. Cf. also (25). (25) a. That way is a manner/a method/a fashion. b. ?That manner/?That method/?That fashion is a way. The remaining question concerning Lexical Selections is why among the four types of BNAs, only temporal BNAs allow much wider variety of head nouns than the other three types, as shown in (2). One might regard this fact as exceptions to the generalization (21). On the con- trary, the facts in (2) pertain to (21), because, as shown in (26), possi- ble interpretations of null temporal N, in fact, range more widely than is shown in (19a), and cover all calendrical units corresponding to a variety of head nouns of temporal BNAs in (2). (26) a. [The time/minutes between 5:30 to 6:00] is a nice period of time for taking a photograph. b. [The time/hours between 1:00 to 6:00] is a nice period of time for taking a photograph. c. [The time/days between August 3rd to Sept 5th] is a nice period of time. d. [The time/weeks between the first week of April to the second week of May] is the busiest period of time. 366 ENGLISH LINGUISTICS, VOL. 16, NO. 2 (1999)

e. [The time/months between May to August] is a nice period of time. f. [The time/years between 1999 to 2005] is a nice period of time. Then, one might suspect that our generalization (20) is incorrect. However, the fact in (2) is, in fact, what is predicted by (20). Note that among the four adverbial theta-roles, Temp is the only concept that can be represented as units; all the head nouns of BNAs except time in (2) are units of time. Since the semantics of a certain unit of time is not more contentful than that of time itself, with a unit of time being time itself and not a type of time like war, nouns denoting units of time are given the same rank as the noun time in the light of seman- tic contentfulness. Besides, with there being no difference in the de- gree of semantic contentfulness among each unit of time, once the noun time is selected as a head of BNAs based on (20), any noun de- noting a unit of time can be selected as well.10 In contrast, since the notions of place, direction and manner cannot be represented as units, (21), based on (20), chooses nouns place (Loc), way (Dir) and way (Man) uniquely as heads of each type of BNAs.

4.2. Typological Restrictions Although our analysis assumes that null Ns in "PP-arguments" and the corresponding abstract nouns in BNAs share non-overt nouns as their common sources, the mere presence of a non-overt noun is not enough for the generation of BNAs. While a PP-argument occurs when a language has a relevant type of null N, self-theta-marking by an abstract noun is achieved in much looser structural relation between theta-assigners and theta-assignees than in the usual cases of theta-

10 Since our analysis allows the abstract noun TIME to be morphologically real- ized as any arbitrary noun denoting a unit of time, we have to guarantee that the morphological realization of TIME at PF and its interpretation at LF accord with each other. As a possible solution to this problem, we can assume that when a unit of time is selected as the head of a temporal BNA, the derivation starts with an abstract noun denoting a unit of time such as DAY, WEEK or YEAR, which are optional realizations of TIME. I owe a reviewer for reminding me of this problem. Cf. Nagai (1986, 1987) for a different view on the choice among the head nouns in temporal BNAs. ANOTHER APPROACH TO BARE-NP ADVERBIALS AS NOMINALS 367 marking. Accordingly, the adoption of a particular type of abstract noun depends on whether the theta-marking in the looser structural condition is allowed for the relevant type of non-overt noun. This consideration leads us to the generalization (27). (27) Possible types of PP-arguments in a language include possi- ble types of BNAs. We regarded the examples in (8) and (10)-(11) in Section 3.1 as evi- dence to show that English has nouns denoting time and place in its repository of non-overt nouns. Similarly, examples in (28) with PP- arguments indicate the existence of a non-overt noun denoting manner in English. (28) a. With a broom was Grandma's favorite way to scare away racoons. (Conway (1997: 61, fn. 7)) b. They considered on foot to be too slow. (Jaworska (1986: 360)) Also, in our analysis, Chametzky's (1985: 33) examples in (29) show the presence of the non-overt noun denoting direction. He points out that (29a) containing a Path PP is better than (29b) with a Goal PP or (29c) with a Source PP. (29) a. Through the Panama Canal seems like a good route for our boat trip. b. ?*To Jamaica seems like a good goal for our boat trip. c. ?*From Iowa City makes an odd origin for our boat trip. Since the direction can. be defined by specifying a particular route, we assume that the Path PP subject in (29a) is an NP headed by a null N denoting direction. It turns out that the four types of BNAs correspond exactly to the possible types of null Ns heading the "PP-arguments" in English in accordance with the generalization (27). The unacceptability of the examples (29b, c) (and similarly (30a-c) for that matter) is also com- patible with (27), since those examples suggest the absence of non- overt nouns denoting goal, source, benefactive, accompaniment and reason in English, and English obviously lacks corresponding BNAs. (30) a. *I regard for Dale as a terrific benefactee(?)/recipient(?)/ beneficiary(?) of this purchase. (Chametzky's (13c)) b. *I regard with John as a good companion. c. *I regard for the delay of trains as a good excuse for being late. While English has four non-overt nouns which can occur as self-theta- 368 ENGLISH LINGUISTICS, VOL. 16, NO. 2 (1999) marking abstract nouns, Spanish PP-argument constructions show that in addition to the four non-overt nouns corresponding to English coun- terparts, as shown in (31a-d), it also has a non-overt noun denoting accompaniment, as shown in (31e). 31) a. Despues de las 5:00 es el mejor tiempo para after 5:00 o'clock is the best time for estudiar el espanol. (temporal PP) study Spanish '(Lit.) After 5:00 is the best time for studying Spanish.' b. Cerca de aqui esta bien. (locative PP) near here is good '(Lit.) Near here is good.' c. Entre Reforms y Cuauhtemoc es 1a mejor para it a between Reforms and Cuauhtemoc is the best for go to la pasteleria. (PP of direction) the cake-shop '(Lit.) Between Reforms Street and Cuauhtemoc Street is the best for going to the cake-shop.' d. Con chaplinos es 1a mejor manes de beber cerveza. with cricket-snack is the best way to drink beer (PP of manner) '(Lit.) With cricket-snack is the best way to drink beer.' e. Con ninos no es recommendable, si visitan With children not is recommendable if (you) visit museos en Italia. (PP of accompaniment) museums in Italy '(Lit.) With children is not recommendable, if you visit museums in Italy.' However, as far as BNAs with internal structures of DP are concerned, Spanish has only temporal BNAs, as shown by the following examples. (32) a. Fui a Oaxaca (en) esta manana/ (en) el ultimo (I) went to Oaxaca (on) this morning/(on) the last sabado/ (en) aquel dia/(en) esta semana/(en) este Saturday/(on) that day/(in) this week (in) this verano. (temporal) winter ' (Lit.) I went to Oaxaca (on) this morning/(on) the last Saturday/(on) that day/(in) this week/(in) this winter.' ANOTHER APPROACH TO BARE-NP ADVERBIALS AS NOMINALS 369

b. He vivido *(en) todo los lugares donde vivio (I) have lived (in) all the places where lived Maria. (locative) Mary '(Lit.) I have lived (in) all the places where Mary lived.' c. He ido *(a) aquel lugar. (direction) (I) have gone to that place '(Lit.) I have gone (to) that place.' d. Lo hice *(de) esta manera/*(con) este metodo/*(de) (I) did (in) this manner/(with) this method/(in) este modo. (manner) this mode '(Lit.) I did it (in) this manner/(with) this method/(in) this mode.' e. Fui a Oaxaca *(con) ninos. (accompaniment) (I) went to Oaxaca with children '(Lit.) I went to Oaxaca (with) children.' Note, however, Spanish also has DP proforms like aqui and aca"for' "here," and alli, alla and ahi for "there" and asi for "so," etc. We re- gard them as BNAs which are produced through PF-merger.11 This situation is still in accordance with (27); Spanish has at least five non- overt nouns, of which four excluding the one denoting accompaniment can take the form of self-theta-marking abstract nouns. Among the four self-theta-marking abstract nouns, only TIME makes BNAs with internal structures of DP. The other three are realized as DP - forms.

11As in Kayne (1993),we assumethat somelexical items can be spelled-out forms of abstract lexical items or of their incorporated forms. In this analysis, pro- forms in English like then and there and deictics like now and here can be generated as BNAs, because we assume that these expressions are DPs produced through the process of Merger in the PF component under PF adjacency (cf. Lasnik (1995)). That is, as have is assumed to be the spelled-out form of D+BE in Kayne (1993), PF-merger produces then from the abstract DP THAT+TIME, there from THAT+PLACE or THAT+DIRECTION, now from THIS+TIME and here from THIS+PLACE or THIS+DIRECTION. These DP proforms also show the properties of BNAs in general. For example, Semantic Extension (based on the generalization (36) in Section 4.3) is also observed in the derivation of these pro- forms. 370 ENGLISH LINGUISTICS, VOL. 16, NO. 2 (1999)

4.3. Semantic Extension Relevant to the fact in (4) is Abney's (1987: 74-75) remark that, as contrasted with other determiners, a and the are functional heads which lack descriptive content and are inseparable from their complements. It is not the case, however, that heads of BNAs accompanied by those determiners always result in unacceptable BNAs. If such structures in- clude adjectival modifiers, the resulting DPs again make acceptable BNAs. While BNAs like the whole day, the other day, the next day, the same day, or the previous day may sound rather idiomatic, this fact is quite general, as shown by the examples in (33). (33) a. They put on the play an evening I well remember. (Suzuki (1987: 46, fn. 14)) b. I must say that it is interesting both ways, .... (Japan Times) c. They get their expertise the old-fashioned way;... (Effective Speech Communication, Glencoe) d. We met the day of the conference. (Quirk et al. (1985:693)) e. I saw her the day she visited our town for the first time. f. We tend to do things the way we've done them before,... (Understanding Psychology, Glencoe) The fact that "modifications" by meaningful determiners in (4b-e) and by additional modifiers in (33a-f) remedy the situation in (4a) suggests that the restriction in question is of semantic sort. As Tominaga (1988:74) points out, this property of BNAs is reminis- cent of similar facts of cognate object constructions in (34)-(35).12 (34) a. Mary laughed a hearty laugh. b. *Mary laughed a laugh. (35) a. John slept a sound sleep. b. *John slept a sleep. To describe this fact of cognate object constructions, Nakagawa (1997) proposes a generalization (36). (36) If a theta-assigner contains the semantic content of a theta- assignee and the theta-assignee is not specialized, the theta- assignee cannot be realized as an independent element. (Nakagawa's (79))

12 In fact, Tominaga (1988:74) claims that cognate objects are BNAs. ANOTHER APPROACH TO BARE-NP ADVERBIALS AS NOMINALS 371

Interestingly, in our analysis, the condition in (36) holds true in the case of theta-marking inside BNAs. Due to the self-theta-marking property of head Ns of BNAs, the head is a theta-assigner and the DP containing the head is a theta-assignee, and the theta-assigner contains the semantic content of the theta-assignee. Hence, according to (36), unless the DP is "specialized," i.e., unless a nominal head of BNAs is accompanied by a or a modifier which are meaningful enough, it cannot be realized as an independent element. Assuming, as above, that determiners a, the and the null determiner lack descrip- tive content, in contrast with other determiners and adjectival modifiers in (4b-e) and (33), (36) correctly excludes (4a) among the examples in (4).13 This analysis also predicts that if a definite and an indefinite article have descriptive content in some language, that language adopts the ex- pressions corresponding to "the N" and "a N" as BNAs. In Spanish, definite and indefinite articles can be used as pronominals without com- plements, as shown in (37). (37) a. Flor, voy a guardar las blusas blancas en el Flor (I'm) going to put the blouses white in the cajon de arriba y las de color en el de abajo. drawer above and the of color in the under '(Lit.) Flor, I'm going to put the white blouses in the

13 We have to regard the inseparability of determiners from their complements only as one of the diagnostics of determiners' lacking descriptive content. Other- wise, we would incorrectly predict that the BNAs of the form "every N" are un- acceptable, because the determiner every is inseparable from its complement, as attested by the fact in (i): (i) a. Every student failed. b. *Every failed. Examples in (ii) show another idiosyncratic property of the determiner every. Though nouns period, lecture and game do not refer to calendrical units, they can appear as heads of BNAs when they are "modified" by every. Provided that the BNAs in (ii) are well-formed, the determiner every seems to have a special property of rendering the semantics of non-calendrical nouns calendrical in appropriate con- texts. For example, in the context of school life, lecture can be a unit of time, and every lecture becomes a synonym of "always." (ii) He takes risks (in) every period/lecture/game. (Suzuki (1987:41)) It is beyond the scope of this paper to account for these idiosyncratic characteristics of every. 372 ENGLISH LINGUISTICS, VOL. 16, NO. 2 (1999)

above drawer and the of color in the below.' (Linguaphone: Curso de espanol hispanoamericano: 159) b. Fernando, tienes nietos. Una vive en Puerto Fernando, (you) have grandchildren One lives in Puerto Rico y el otro esta aqui en Mexico. Rico and the other is here in Mexico '(Lit.) Fernando, you have grandchildren. One lives in Puerto Rico and the other is here in Mexico.' (Spanish Conversation, NHK TV, Jan, 1999:62) If the separability of determiners from their complements, i.e. the pro- nominal use of las and el in (37a) and that of una in (37b), suggests that the determiners have descriptive content, the examples in (37) sug- gest that definite and indefinite articles in Spanish have descriptive con- tent. Hence, it is predicted that appropriate heads accompanied solely by definite or indefinite articles can make well-formed BNAs in Span- ish. This prediction is borne out, because Spanish has BNAs like el jueves(=this Thursday), un dia (=one day), una noche (=one night), etc.

4.4. Semantic Inclusion As mentioned in Section 2, temporal BNAs can only subsume the semantics of prepositions at, on, in and for but not of, say, before or after. One can assume that in temporal PPs like at that time, on Mon- day, in the morning or for hours, the prepositions denote a point or in- terval of time. However, it is also possible to regard those preposi- tions as semantically vacuous Case-markers, as in Enc (1987:640, fn. 7), and that a point or interval of time is denoted by the objects of prepositions.14 The fact in (38) observed by Jaworska (1986) seems to show that the temporal preposition in, in its normal non-contrastive use, is a semantically vacuous Case-marker. (38) a. ?In Poland was safe. b. Poland was safe. c. In Poland was safe but outside Poland was most danger- ous.

14 Although Enc (1987:640, fn. 7) refers to only temporal on and temporal in as Case-assigners, the same analysis seems to apply naturally to other prepositions mentioned in the text. ANOTHER APPROACH TO BARE-NP ADVERBIALS AS NOMINALS 373

Jaworska (1986) points out that PP-argument constructions are allowed only when the presence of prepositions is justified semantically. That is, P+NP is allowed to occur as an argument only in the cases where the NP per se cannot have the same semantics as P+NP. Accordingly, the lower acceptability of (38a) in comparison with (38b) indicates that the locative NP Poland per se has the same semantics as that of in Po- land. The PP argument in Poland is allowed to occur only in contexts like (38c) where the presence of the preposition in is justified because of its additional contrastive meaning.15 Since it is obvious that the incompatibility with PP-argument struc- tures represented in (38a) is observed with all the other prepositions whose semantics is subsumed by BNAs (such as temporal Ps at, on, in and for, locative Ps at and in and Ps of manner and direction in), we can assume that those prepositions are all semantically vacuous Case-

15 It seems that a principle of economy of representation accounts for the lower acceptability of (38a) in contrast to (38b); if (38a) and (38b) are synonymous, the structure of the former with null-headed NP subjects has a representation less eco- nomical than the structure of the latter with a simple NP subject. Conway (1997) applies a semantic analysis to this fact and points out various interesting features of constructions like (38a). Since Conway claims that null N heads carry with them null definite determiners, the unacceptability of (ia, b) with null-headed NP subjects is attributed to that of (iia, b) with simple lexically-headed definite NP subjects. (i) a. *In San Jose is a good place to hold a conference. b.??/?*In January is fine with me. (ii) a. *the place in San Jose b. ?*the time in January On the other hand, the acceptability of examples like (iiia, b) is attributed to that of (iva, b). (iii) a. Under the chair is a nice place for the cat to sleep. b. Before the war is a distant memory. (iv) a. the place under the chair b. the time before the war As for the reason why PPs like in San Jose and in January are incompatible with defi- nite determiners, Conway claims that to be definite, locative PPs and temporal PPs must represent a bounded area and bounded span of time, respectively. In this re- spect, PPs like in San Jose and in January only represent a point in space and time and are incompatible with definite determiners. If Conway is on the right track in claiming that null N heads carry with them null definite determiners, we have to assume that it is a property only of null Ns occurring as nominal heads of "PP-argu- ments" and not that of abstract nouns in BNAs. As we proposed in Section 3.3, Conways' null Ns and the abstract nominal heads of BNAs share common sources in our analysis. However, it is obvious that temporal BNAs can refer to a point in time, as shown by the examples in (2). 374 ENGLISH LINGUISTICS, VOL. 16, NO. 2 (1999) markers. Now, our account of Semantic Inclusion goes as follows. BNAs can occur as a result of Case feature-elimination due to self- theta-marking property of their head nouns. If so, with prepositions being "Case-markers," the semantics of BNAs cannot include the se- mantics of any prepositions whatsoever. The reason why BNAs look as if they exclusively subsume the semantics only of prepositions at, on, in and for is that these prepositions are mere Case-markers which are semantically vacuous.

5. Consequences In this section, we point out consequences following from our analy- sis of BNAs.

5.1. Non-Wh-Adverbial Relative Our analysis covers an interesting consequence of Larson's (1985) analysis. Showing examples in (39), Larson points out that the possi- ble head nouns for non-wh adverbial relative without are precisely the nouns that can appear as heads of BNAs. (39) a. the month b. the day {(that) you traveled to France t c. the year d. *the vacation (for you) to travel to France t} e. *the occasion} (Larson's (41)) Larson proposes an analysis where [+F] feature of month, day and year is "absorbed" by the following null operator and transmitted through the operator chain to its trace. As a result, the trace is successfully Case-marked and licensed as a variable, and (39a-c) are generated. On the other hand, this process, allegedly, cannot occur in the deriva- tions of (39d, e), due to the lack of [+F] feature with vacation and occasion. In contrast, we can propose that month, day and year are allowed as head nouns in (39) because the abstract noun TIME, of which they are morphological realizations, has (self-)theta-marking property to trans- mit to the tail of the operator chain. We need not rely on a process like absorption because the acceptability of examples like (40) indepen- dently shows that the theta-marking property of the head of a is transmitted to the tail through the A-bar chain; otherwise, ANOTHER APPROACH TO BARE-NP ADVERBIALS AS NOMINALS 375

Susan in (40) cannot get a theta-role from Adj foolish.16 (40) Mary considers John foolish, as Bill considers Susan t. (Stowell (1987)) On the other hand, (39d, e) are excluded because the null-operator chains lack adverbial theta-roles, due to our assumption that the heads of relative clauses, i.e., the vacation and the occasion are not morpho- logical realizations of the self-theta-marking abstract noun TIME. Emonds' (1987:625, fn. 11) example in (41) (attributed to a reader for Linguistic Inquiry) also favors our analysis rather than [+F] analy- sis. (41) I live in the same places he lives t. Since the NP the same places is Case-marked by in in the main clause, [+F] feature of place is not activated in Larson's analysis. On the other hand, the following A-bar chain headed by a null operator must absorb the [+F] feature in order to license its trace as a variable. Since these two demands contradict with each other, [+F] analysis needs more elaborated device to generate (41). Common nouns like soldier can occur either as an argument or as a , and they take an external argument only when they occur as predicates. However, as shown by examples like (42), it is not the case that when they occur as arguments they lose the property of assigning external theta-roles. That property is only suppressed in such cases. (42) John looked like a soldier, which indeed he was t. In the main clause, the DP a soldier is an argument of the preposition like (or of the complex verb look like) and does not take any argument. However, within the relative clause, once the DP is reconstructed at the post-be position through the A-bar chain headed by which (assum- ing the movement analysis of which as a relative-operator, as in Stowell (1987)), the noun soldier functions as a predicate taking he as its exter- nal argument. Otherwise, the subject he in the relative clause would be void of a theta-role. (42) shows that the theta-marking property of a head of a relative-clause can be transmitted to the tail of the operator

16 We adopt Stowell's (1987:50) suggestion that the derivations of examples like (40) involve movement of empty AP operators. 376 ENGLISH LINGUISTICS, VOL. 16, NO. 2 (1999)

chain, though it is not activated in the head itself. The same analysis applies to (41). In the main clause, self-theta-marking property of the head of the DP the same places is not activated. However, once the DP is reconstructed through the A-bar chain at the tail position, that property of the head noun places is reactivated, and satisfies the chain condition.

5.2. -ly By regarding -ly adverbs as another member of adverbials theta- marked by their own heads, our analysis gives an answer to a tradition- al question of why, as pointed out by Jackendoff (1977), adverbs do not take complements.17 We assume that English has a non-overt manner , which takes the form of either a null adverb or an abstract adverb MANNER. As a null adverb, it takes an AdjP-complement, and satisfies its morpho- logical requirement as an affix, by attracting Adjectival head at LF. As an abstract adverb MANNER, on the other hand, its non-affixal status allows it not to take a complement AP and in that case, it assigns Manner-role to its own maximal projection, i.e. AdvP. Provided that the observations similar to (20) and (21) should be applied here, the abstract adverb MANNER is spelled out as -ly. The fact that the adverbial head -ly always follows an is compatible with its PF- affixal nature and the generalization (36) in Section 4.3. The structure of depictive predicates in (43a) represents the derivation involving Adj-to-Adv movement and that of -ly adverbs in (43b) the derivation involving self-theta-marking. The structure (43a) accounts for such adverbial status of depictive predicates as pointed out by Maruta (1995). (43) a. They married [AdvP [Adv MANNER] [AdjP young]] b. He drove the car [AdvP [AdjP careful] [Adv -ly]]

5.3. BNAs in Japanese It seems that in Japanese, BNAs are restricted only to temporal DPs. Thus, instead of the temporal postpositional phrases like DP-ni, tem-

17 Iowe this point to SatoshiOku. ANOTHER APPROACH TO BARE-NP ADVERBIALS AS NOMINALS 377 poral DPs can appear without the postposition ni and function as adverbials in the following examples: (44) Sono toki/Sono syunkan/Sono hi/Sono syuu/Sono tukil that time/that moment/that day/that week/that month/ Sono tosi kimi-ga umareta. that year you-Nom be-born-Past 'You were born that time/that moment/that day/that week/ that month/that year.' However, if, as we have claimed in fn. 11, combinations of abstract de- terminers and abstract head nouns can be spelled out as BNAs through PF-merger, Japanese turns out to have more instances of BNAs than is illustrated only by temporal BNAs in (44). In fact, Japanese has loca- tive DP proforms like koko, soko, asoko and doko whose seman- tic contents are THIS+PLACE, THAT(middle distance)+PLACE, THAT (long distance)+PLACE and WHICH+PLACE, respectively. As directional DP proforms it has kotti, sotti, atti and dotti derived, respectively, from THIS+DIRECTION, THAT(middle distance)+ DIRECTION, THAT(long distance)+DIRECTION and WHICH+ DIRECTION. As DP proforms of manner, it has koo, soo, as and doo derived, respectively, from THIS+MANNER, THAT(middle dis- tance)+MANNER, THAT(long distance)+MANNER and WHICH+ MANNER. For example, the nominal head of locative postpositional phrase koko-ni, that of directional postpositional phrase sotti-ni and that of a postpositional phrase of manner koo(-no yoo)-ni can occur alone as BNAs, as shown by the following examples: (45) a. Koko suwaroo. (locative) here let's-sit 'Let's sit here.' b. John-ga sotti itta. (direction) John-Nom there go-Past 'John went there.' c. John-ga e-o koo kaita. (manner) John-Nom picture-Acc this way draw-Past 'John drew a picture this way.' Note also that these DP proforms occur as arguments, as well, as shown in (46). (46) a. Koko-ga ii. here-Nom good '(Lit.) Here is good.' 378 ENGLISH LINGUISTICS, VOL. 16, NO. 2 (1999)

b. Sotti-ga ii. '(Lit.) There is good.' c. Koo-ga ii. '(Lit.) This way is good.' In sum, we claim that while temporal BNAs are generated as DPs with internal structures, other three types of BNAs are realized as DP proforms in Japanese. It is assumed, in this analysis, that Japanese has null Ns denoting time, place, direction and manner, and this assumption is supported by the existence of the corresponding "PP- arguments" in (47). (47) a. Asita-kara-ga tanosimi da. tomorrow-from-Nom delight be '(Lit.) From tomorrow is a delight.' b. Dokusyo-wa furo-de-ga itiban tanosii. reading-Top bath-in-Nom first pleasant '(Lit.) As far as reading is concerned, in the bath is the most pleasant.' c. Kono miti-kara-ga tikai. this road-through-Nom close '(Lit.) Through this road is close.' d. Tuukin-wa toho-de-ga risooteki da. commuting-Top on foot-with-Nom ideal be '(Lit.) As far as commuting is concerned, on foot is ideal.'

6. Conclusion We have shown that Lexical Restrictions and Typological Restric- tions of BNAs follow from our assumptions relevant to non-overt nouns and that Semantic Extension and Semantic Inclusion from our assumptions relevant to licensing of adverbials. While this result indi- cates the plausibility of those assumptions, it also shows that those seemingly idiosyncratic properties of BNAs are, in fact, natural con- sequences reflecting their derivations. ANOTHER APPROACH TO BARE-NP ADVERBIALS AS NOMINALS 379

REFERENCES

Abney, Steven P. (1987) The English Noun Phrase in It3 sentential Aspect, Doctoral dissertation, MIT. Bresnan, Joan and Jane Grimshaw (1978)“The of Free Relatives in English,”Linguistic Inquiry 9, 331-391. Chametzky, Robert (1985)“NPs or Arguments: Exocentricity vs. Predication,” CLS 21, 26-39. Chomsky, Noam (1995) The , MIT Press, Cambridge, MA. Collins, Chris (1997) Local Economy, MIT Press, Cambridge, MA. Conway, Laura (1997)“NPs in Disguise,”University of Connecticut Working Papers in Linguistics 8 (Is the Logic Clear?: Papers in Honor of Howard Lasnik), 55-87. Emonds, Joseph E. (1987)“The Invisible Category Principle,”Linguistic In-

quiry 18, 613-632. Enc, Murvet (1987)“Anchoring Conditions for Tense,”Linguistic Inquiry 18, 633-657. Fukui, Naoki and Yuji Takano (1998)“Symmetry in Syntax: Merge and De- merge,”Journal of East Asian Linguistics, 27-86. Jackendoff, Ray (1977) X-Bar Syntax: A Study of Phrase Structure, MIT Press, Cambridge, MA. Jaworska, Ewa (1986)“Prepositional Phrases as Subjects and Objects,”Journal of Linguistics 22, 355-374. Kayne, Richard S. (1993)“Toward a Modular Theory of Auxiliary Selection,” Studia Linguistica 47, 3-31. Kobayashi, Keiichiro (1987)“A Note on Bare-NP Adverbs,”English Linguis- tics 4, 336-341.

Kobayashi, Keiichiro (1999)“Bare-NP Adverbials Revisited,”Kagaku/Ningen

(Journal of Science and Humanities) 28, 93-122, Kanto Gakuin University. Larson, Richard K. (1985)“Bare-NP Adverbs,”Linguistic Inquiry 16, 595-621. Lasnik, Howard (1995)“Verbal : Syntactic Structures Meets the Minimalist Program,”Evolution and Revolution in Linguistic Theory: Essays in Honor of Carlos Otero, ed. by Paula Kempchinsky and Hector Campos, 251-275, Georgetown University Press, Washington D.C. Longobardi, Giuseppe (1994)“Reference and Proper Names,”Linguistic In-

quiry 25, 609-665. Maruta, Tadao (1995)“The Semantics of Depictives,”English Linguistics 12, 125-146. McCawley, James D. (1986)“Adverbial NP's: Bare or Clad in See-Through Garb,”Metropolitan Linguistics 6, 1-13, Linguistics Circle of Tokyo Metro- politan University. Nagai, Tomotaka (1986)“Toki-o Arawasu Fukushi-teki Meishi-ku (Adverbial Noun Phrases Expressing Time) (1),”Eigo Kyoiku 35, 10, 67-69, Taishu- kan, Tokyo. 380 ENGLISH LINGUISTICS, VOL. 16, NO. 2 (1999)

Nagai, Tomotaka (1987)“Toki-o Arawasu Fukushi-teki Meishi-ku (Adverbial Noun Phrases Expressing Time) (2),”Eigo Kyoiku 35, 11, 64-66, Taishu- kan, Tokyo. Nakagawa, Naoshi (1997)“Tough as Null Derivational Mor-

phemes,”English Linguistics 14, 230-250. Neelman, Ad (1997)“PP-Complements,”Natural Language and Linguistic Theory 15, 89-137. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik (1985) A Comprehensive Grammar of the English Language, Longman, London. Rothstein, Susan D. (1983) The Syntactic Forms of Predication, Doctoral dis- sertation, MIT. [Published by the Indiana University Linguistics Club, 1985] Stowell, Tim (1987)“As So Not So As,”ms., University of California, Los Angeles. Stroik, Thomas (1992)“On the Distribution of Temporal and Locative Adver- bials,”Linguistic Review 9, 267-284. Suzuki, Hidekazu (1987)“Meishiku-no Fukushiteki Youhou ni kansuru Ichi- kousatsu: Larson (1985)‘Hadaka Meishiku Fukushi’o Megutte (A Note on Adverbial Usage of Noun Phrases: Concerning Larson's (1985) Bare- NP Adverbs),”Gengo-Bunka Ronshu 22, 29-46, University of Tsukuba. Takahashi, Daiko (1993)“On Antecedent Contained Deletion,”ms., Univer- sity of Connecticut, Storrs. Takano, Yuji (1996) Movement and Parametric Variation in Syntax, Doctoral dissertation, University of California, Irvine. Tominaga, Hideo (1988)“Told-no Fukushi-teki Meishi-ku (Temporal Adverbial Noun Phrases),”TAM Siron-shu 1, 63-76.

1-17-13-506 Katakura Kanagawa-ku Yokohama-shi Kanagawa 221-0865 e-mail:[email protected]