Abstract
In this chapter, we will see some consequences of the Three-Layered Case Checking Hypothesis with respect to Accusative Case checking. Recall that our theory requires the presence of Tns to check off the Accusative Case feature of the verb in case of Accusative Case checking. It is not easy to find evidence that Tns is implicated in the series of processes for Accusative Case checking, since Tns is always there in the ordinary clauses. If we look at reduced clauses, however, we are led to conclude that structural impoverishment is responsible for Case absorption.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
Notes
It is Nominative Case if no special embedding is involved. It is Accusative in the ECM context. In both situations, the passive verb is not responsible for the shape of Case. We are also abstracting away from impersonal passives in this section. See section 3.2.4.1 for a brief discussion.
See Siewierska (1984) and Haspelmath (1990) for passive morphology. See also Dobrovie-Sorin (1993) for two kinds of the passive constructions in Romanian.
It takes the form of -In after vowels and /11, and -II otherwise.
Bittner (1994) proposes to account for the West Greenlandic passive by recognizing the biclausal structure. It is hoped that our account to be presented below will carry over to Ergative languages like West Greenlandic.
Cf. also Baker (1988) and Roberts (1987). Precursors of their idea can be found in Belletti (1982) and Jaeggli (1986a).
Irrelevant positions are also omitted.
Cf. Hoekstra (1984, 1986) and Campbell (1989) for a similar idea. Hoekstra, however, is more concerned with auxiliary choice of the kind that Burzio (1986) discusses in relation to the unaccusativity hypothesis. We will discuss Aux selection in section 3.2.4. Cowper (1989) also proposes a similar analysis, which assumes external theta role transmission. The difference is that for her, -EN is an affix which simply removes the external theta role and the Accusative Case feature of participles. In the perfect, the Accusative Case feature is assigned by have to the participle, which in turn assigns it to the direct object. See also Noonan (1992, 1993) for a somewhat different implementation of the same idea.
These recent ideas can be traced back to Bach (1967) and Benveniste (1966). Cf. also Tremblay (1991).
We will touch on the question of Aux selection in section 3.2.4.
Kayne’s (1993) proposal is that the participial construction has the skeletal structure (i). See also Schafer (1994) for evidence from Breton for the Kayne-Freeze hypothesis. In section 3.2.3.2, we will take up Schafer’s analysis, which closely parallels the structure (i).
For an Aspect Phrase, see Hendrick (1991), McClure (1993), Ramchand (1993), and Travis (1991b), among others.
Cf. also Kayne (1989).
So-called Long Head Movement (LHM) in the perfect, discussed extensively in Lema and Rivero (1990, 1991) and Rivero (1991, 1994a), involves adjunction of the participial verb to the auxiliary, according to the Minimalist proposal by Wilder and ~avar (1994). This may be another instance, but discussion of LHM will take us too far afield. Cf. also Boskovib (1995).
It goes beyond the scope of this book to provide an account of the simple copular constructions. Our suggestion is that Accusative Case checking in the possessive construction takes place within a projection whose head is incorporated into the copula. Note that the have-type copula does not always check Accusative if the following French construction should be analyzed as involving LF movement of the postverbal element to the expletive pronoun. (i)Il y a deux enfants dans l’auto.it there has two children in the car There are two children in the car.’ Freeze (1992, 567) Alternatively, we might say that the copula which is to become the possessive have has an Accusative Case feature, in contrast to the perfect have, since the two need not be indentical in every respect, as emphasized in the text. The second possibility weakens Freeze’s hypothesis, though.
We are abstracting away from the effects of head movement again.
This ordering is forced by Strict Cyclicity when both subject raising and object raising take place in overt syntax. We have seen in Chapter 2 that cooccurrence of subject raising and object raising at LF creates a Minimality problem because of the disappearance of Agr chains at LF.
See Bures (1993), Déprez (1989), Mahajan (1990), and Wyngaerd (1989), among others, for object shift as movement into Spec of Agr-oP. See also Holmberg (1986), Vikner (1990), and Webelhuth (1989) on object shift.
For our immediate purposes, the question whether Dutch is head-final or head-initial does not matter. If the Dutch VP is head-final, (3.13a) shows the underlying order. See Zwart (1993b, 1994a) for a detailed discussion of the head direction in Dutch.The text account of object shift in Dutch presupposes the head-final nature of Dutch. For the head-initial account of object shift, see Zwart (1993b, 1994a,b). In the latter account, we have to assume additional leftward movement of the object from Spec of Agr-oP, in order to be consistent with the current set of assumptions, as Zwart (1994b) notes.
This is the point overlooked by Bobaljik (1994, 1995) when he discusses the Minimality problem posed by the periphrastic constructions in SOV Germanic languages. Verb raising in the sense of Evers (1975) is characteristic of SOV Germanic languages and makes object shift possible in the auxiliary constructions.
This is the Extended Projection Principle effect.
They refer to Hoekstra (1984).
We abstract away from optionality of agreement in some dialects of French and with the first person and second person elides in Italian. See Burzio (1986, 59) for the latter optionality.
Participle agreement takes place in unaccusative clauses as well, indicating that Case checking and participle agreement are independent of each other. Crucial for us is that agreement is induced by the Spec-head relation in the participial AgrP. Note also that the fact that the clitic is attached to the finite auxiliary cannot be taken as evidence for the Accusative Have Hypothesis. There are languages such as Franco-Provençal Mayne 1991) and Brazilian Portuguese (Bianchi and Figueiredo Silva 1994) where clitics are attached to the past participle. The final landing site of clitics is governed by something different from structural Case itself.
Branigan (1992), Friedemann and Siloni (1993), and Sportiche (1990) share this assumption. We will come back to Branigan and Sportiche’s argument in the next chapter.
Kayne (1989) proposes adjunction to AgrP for participle agreement triggered by wh-movement, but not for participle agreement induced by cliticization. In Chapter 4, we will discuss adjunction to AgrP accompanied by wh-movement.
Branigan (1992) proposes to account for the participle agreement by saying that clitic movement takes place after subject raising. He claims that this derivation does not violate Strict Cycle, because adjunction is exempted from Strict Cycle in Chomsky’s (1993) account. We will not adopt this proposal, since it does not generalize to the cases to be discussed in Chapter 4. Furthermore, there is a theoretical reason to suppose that adjunction is also subject to the condition that derives Strict Cycle. See Watanabe (1995a) for adjunction and Strict Cycle.
The ambivalent status of adjoined positions is first noted by Tada (1993) and Um (1993).
Duffield (1995) discusses other differences between the progressive particle and the perfect particle. Let us point out here that the Welsh perfect and the Irish progressive allow the passive interpretation, in addition to the ordinary active reading.(i) a. Y mae tai newydd wedi eu hadeiladu yng Nghaerdydd Prt is houses new Perf.Prt 3p1. build-VN in Cardiff ‘New houses have been built in Cardiff.’ (Welsh) Sproat (1985, 191) b. Ta tithe tira a dt6gail acu i nDoire are houses new 3p1. build-VN at-them in Derry ‘New houses are being built by them in Derry.’ (Irish) McCloskey (1983, 30) For these constructions, we assume that Asp is not doing follow-up checking. It is interesting to note that the progressive particle is missing in the Irish construction in (ib). Cf. the Irish active progressive construction in (ii). (ii) Ta mé ag t6gail teach tir am I Prog.Prt. build-VN house new ‘I am building a new house.’ (Irish)
Breton is exceptional among the Celtic languages in this respect. We will turn to Breton shortly.
If the copular verb is used at all. Other verbs such as become and go are also used in the passive in some languages.
Lasnik attributes the original observation to Quirk et al. (1972).
But not the auxiliaries have and be, crucially for his purposes. These auxiliary verbs are generated together with features, as is the case with the French verbs. See the discussion in his paper.
We ignore the slight marginality of this example, noted by Lasnik.
If Agr lacks phonological features as we have assumed in Chapter 1, the elided part need not be Agr(-o)P.
We assume deletion in the PF component, as suggested by Chomsky (1993).
Note also that ellipsis is possible in the passive as in the perfect. (i) Gary should have been paid better, and Mary should have been (paid better), too. Baker, Johnson, Roberts (1989, 244) This is consistent with our analysis of the passive.
An alternative to the Affix Hopping analysis is to play with a hierarchical organization of features. Suppose that the finite form of a verb has the structure [Agrn,… [Tns [V0]]], the infinitival and the past participle forms [Agra,… [V0]],and the progressive form [[Agra,… [Asp [V0]]]]. Ellipsis can take place, if the antecedent contains a subcomponent that is identical to the head of the elided VP, ignoring Agr features. The past tense form contains the substructure [V°], so that it can serve as antecendent of the infinitival VP, but not of the progressive VP which has Asp. According to this hypothesis, it is not necessary to have optional ordering between Affix Hopping and ellipsis in accounting for (3.34e) vs. (3.340. The fact that have and be require strict identity, as in (i), will be attributed to lack of featural organization at the relevant level. (i) John was here, and Mary will *(be here), too. The copula be has the structure [Agra,… [Past, V0]] in the past tense and [Agra,… V0] in the infinitive, for example. Notice that there is no subcomponent that can be the antecedent of the deleted VP. Comparison between this alternative and Lasnik’s proposal amounts to the trade off between Affix Hopping vs. featural organization, it seems. We put aside exploration of this issue, because it will take us too far afield. We will propose, however, that featureal organization is necessary in another domain in section 4.3.2.1 in the next chapter. For a hierarchical organization of features in Romance clitics, see Bonet (1991).
Schafer labels Agr2 as Agr-o, though she does not discuss Accusative Case checking in her paper. If she intends Spec of the lower AgrP to be the locus of Accusative Case checking, it is consistent with our claim.
It is worth noting that the progressive in Breton is expressed by the periphrastic construction containing bezan together with a particle o (Hendrick 1991, Press 1986). (i) Da Gemper e oan o vont to Quimper Prt was-lsg Prog.Prt go ‘I was going to Quimper.’ Hendrick (1991, 174)
The have-type copula kaoud is also used for the possessive construction. See Schafer (1994) for discussion.
Hendrick (1994) proposes the structure (i) for the perfect in Breton. (i) IVP bezan [AgrP Agr [AspP Asp VP]]] The structure (i) is close to our (3.38), except for the absence of AgrP between AspP and VP. Other differences are: (i) that Hendrick claims that the lower verb incorporates into Asp, instead of Asp incorporating into bezan, and (ii) that Hendrick does not assume VP-internal subject.
Hendrick (1994) also uses the agreement pattern of kaoud to motivate AgrP in question.
See McCloskey and Hale (1984) for Irish, and Hendrick (1988, 1994) and Rouveret (1991) for Welsh. Complementarity between rich verbal agreement and overt subjects breaks down for pronouns in Welsh. See the references cited for some discussion.
Hendrick (1994) also proposes the same analysis of the exceptional agreement behavior of kaoud.
See Rapoport (1987) and Shlonsky (1987) for the Hebrew copular construction in its predicative, locational, and possessive uses. Hebrew lacks the have-type copula, a point which we turn to below.
This is the reading that we are now concerned with, as Tali Siloni (personal communication) informs me. Perfect is also possible with passive verbs.
This possibility is limited to pronouns, according to Friedemann and Siloni (1993). Movement of the pronominal object may involve adjunction, instead of substitution into Spec, as we have claimed for Romance clitics above. The adjunction option does not affect the main line of argument in the text, however.
Egyptian Arabic, discussed by Diesing and Jelinek (1995), displays essentially the same pattern. Bobaljik (1991) notes that the Slavic languages such as Servo-Croation and Russian use the be-type copula in the participle constructions and that the participle always agrees with the subject even in transitive clauses. Polish seems to show the same pattern, judging from Borsley and Rivero (1994). Cf. also Embick (1995) on Polish. Our analysis is expected to extend to these languages as well.
Kayne posits Tns in order to capture the fact that the be-type auxiliary tends to be avoided in the present tense, which is also true in the Corese dialect. Kayne suggests that this Tns, being defective, must raise into the copula in the present tense, together with DIP°, presumably because the present tense tends to be zero. So, Kayne’s use of Tns is rather different from our Asp. It is not clear whether we need Tns to capture the tendency to avoid the be-type auxiliary in the present tense. Essentially following Kayne’s idea, we may say that incorporation of Asp into the copula is needed in the present tense, giving rise to the have-type copula, in order to supplement the defective nature of the matrix present Tns node, with which the copula checks its Tns feature. Cf. also Giorgi and Pianesi (1991) for the zero nature of the present tense.
See Vikner and Sprouse (1988) for a survey of Romance and Germanic languages. As we briefly mentioned above, Kayne (1993) presents a detailed analysis of variations in Romance languages which is more or less compatible with our framework.
b) is exactly the same as Kayne’s (1993) proposal for the languages which select be for an unaccusative verb.
See also Bobaljik (1993), Hale and Keyser (1993), Laka (1993), and the references cited there for relevant discussions.
We will return to this reasoning more carefully in section 3.4. Here, this brief statement suffices.
One significant problem is that languages vary as to the possibility of the impersonal passive of unergatives. It is impossible in English, for example, as shown in (i). (i) a. Rob ate five times a day. b. *It/there was eaten (by Rob) five times a day. Baker (1988a, 343) Since passivization of transitives shows no parametric variation, this break down of the parallelism between transitives and unergatives looks problematic for the idea that unergative verbs take a phonologically empty cognate object. There is a reason, however, why (ib) is not allowed in English. In the impersonal passive of unergatives, the null object is assumed to be sitting in its original position because the subject position is filled with an expletive. See Dobrovie-Sorin (1993) for some discussion. On the other hand, English disallows the counterpart of (3.50), though we will not discuss why it is impossible. See Lasnik (1992b, 1995) for recent discussion. (ii) *There was killed a man.
We are not committed to the view that being stative or telic is sufficient for the membership of the class of unaccusatives. See in particular Levin and Rappaport Hovav (1995) and Rappaport Hovav and Levin (1992), where it is argued that directed change, but not telicity, is the relevant notion. See also Tenny (1987, 1994) for development of a theory in which aspectual properties play a central role in determining argument structure of verbs.
Thanks to Chris Collins for calling my attention to this class of verbs.
Rappaport Hovav and Levin (1992, note 28) mention that roll verbs in Dutch take hebben ‘have’ in the absence of a directional phrase.
Cf. also Hoffman (1991) and Li (1990a, 1990b).
One of the differences is that the French and Spanish causative resists passivization. Thus, the embedded object cannot become the matrix subject under passivization. (i) illustrates this point in French. (i) a. On a fait construire la maison à Casimiro. one has made construct the house to Casimiro We made Casimiro construct the house.’ b. *La maison a été faite construire à Casimiro. the house has been made construct to Casimiro ‘The house was made to be constructed by Casimiro.’ The Italian counterpart of (ib) is grammatical, as we will see below.
In section 3.4.2, we will turn to a detailed discussion of the Japanese causative, which displays some properties that set it apart from the ordinary ECM construction.
Imbabura Quechua also has the reduced causative, as noted above.
In this study, we will not take up the faire -par construction. This type of the causative, found in Romance (Burzio 1986, Guasti 1992, Kayne 1975, Zubizarreta 1985, 1987) and Bantu languages including Chichewa (Alsina 1992), has the same Case property as the reduced causative. It has been known since Kayne (1975), however, that there are additional restrictions on the direct object of the embedded clause in the faire par construction. Recent studies (Alsina 1992, Guasti 1992) converge on the idea that the embedded object of the faire pa r construction must be an affected object. The reduced causative lacks this restriction.
Knecht (1986, 155) and Zimmer (1976) note that when the embedded verb is ditransitive, the version in which the embedded subject is marked by tarafindan, which is also used to mark the passive agent, is better than the version in which the embedded subject is marked by Dative. Thus, (ib) is preferred to (ia). (i) a. Mtidiir-e mektub-u Hasan-a glister-t-ti-m. director-Dat letter-Acc Hasan-Dat show-caus-past-1 sg ‘I made the director show the letter to Hasan.’ b. Mtidiir tarafindan mektub-u Hasan-a göster-t-ti-m. director byletter-Acc Hasan-Dat show-caul-past-lsg Knecht (1986) also notes, however, that the use of tarcOndan with a simple transitive verb as in (ii) is impossible. (ii) *(Ben) pencere-yi Cengiz tarafindan aç-tir-di-m. I window-Acc Cengiz byopen-caus-past-lsg ‘I had the window opened by Cengiz.’
In French and Spanish, passivization of the causative is not possible. See the references cite above.
The subject of an embedded intransitive also triggers participle agreemen when cliticized
Another possibility is that the embedded oblique subject is an adjunct. The main reason for regarding it as an argument (in Spec of VP) is the binding facts to be reviewed in section 3.4.1. We will come back to this point.
The Accusative Case feature of the embedded verb could be eliminated even if XP in (3.71) were TP, instead of AgrP, given the role of Tns as a follow-up checker.
We omit the oblique external argument of the passive, to which we will return in section 3.4.1.
A in the gloss indicates a noun class of the direct object.
See also Bobaljik (1993) for discussion of noun incorporation. Bobaljik (personal communication) suggests the bare phrase structure treatment of noun incorporation, from which we are borrowing.
Chomsky (1994) suggests that a clitic is both a D° and a DP.
Baker (1995) suggests that the impossibility of causativizing transitive verbs is a general characteristic of polysynthetic languages, except for Inuit.
Verbs of eating are exceptions to the prohibition against embedding of transitive clauses. Verbs of eating behave differently from ordinary transitives under the causative in various languages. See Baker (1988a, 461, note 31) and Alsina and Joshi (1991), among others.
Chinese, Thai, and Vietnamese also share this peculiarity according to Siewierska (1984), and we expect them to be analyzed in the same way as Japanese.
The initial consonant /r/ drops after verbs which end with a consonant.
We will restrict the discussion to the type of direct passive which marks the external argument by niyotte. There is another type of direct passive which marks the external argument by ni. The latter has sometimes been classified with indirect passive which will be introduced below (cf. Kuroda 1979 and Kitagawa and Kuroda 1992 ). See Hoshi (1991, 1994) for an alternative proposal. We will not commit ourselves to the status of the latter type here.
The movement nature of the direct passive can be seen from the possibility of passivizing idiom chunks pointed out by Harada (1977). This is illustrated by the idiom kechi-o tsukeru ‘criticize’ in (i). (i) a. John-go kono bunseki-ni kechi-o tsuke-ta. John-Nom this analysis-Dat KECHI-Acc attach-Past ‘John criticized this analysis.’ b. Kechi-ga kono bunseki-ni tsuke-rare-ta. KECHI-Nom this analysis-Dat attach-Pass-Past This analysis was criticized.’ (ib) is the passive version of (ia). See also Hoshi (1991, 1994), Kitagawa and Kuroda (1992), and Kuroda (1979).
See Howard and Niyekawa-Howard (1976) for a review of the earlier references. Kubo (1990) and Terada (1990) argue that a subset of the instances originally subsumed under the indirect passive display the behavior of the direct passive. Specifically, they argue that NP movement from the original object takes place in these cases. Terada suggests a possessor raising analysis. In what appears to be the indirect passive in Korean, the possessor of the embedded object must be the matrix subject, as noted by Park (1986) and Washio (1993). This construction probably requires the possessor raising analysis, too.
Uniform analyses based on the biclausal structure have been proposed since the earliest stage of generative grammar in such work as K. Hasegawa (1964) and Kuroda (1965). See Howard and Niyekawa-Howard (1976) for discussion of these early proposals. Our account builds on Marantz’s (1985) and N. Hasegawa’s (1988) uniform treatments.
We will modify the position of the external argument in section 3.4.1.1.
In traditional Japanese grammar, the passive morpheme is thought to be derived from the copular verb a r. See Sansom (1928) and Tokieda ( 1950 ). One might say that -r-of the passive morpheme -(r)are is the participial Agr. It is not clear where the final -e in the passive morpheme -(r)are comes from.
We turn to an explanation of Burzio’s generalization in section 3.4.1.2.
Miyagawa (1989) puts forward a lexical version of the uniform analysis, in which the following property of the passive morpheme is stipulated: (i) If -Nate absorbs the Case from the verb that it attaches to, it can optionally assign this absorbed Case. Miyagawa (1989, 172) Under Miyagawa’s approach, a passive verb is attached in the lexicon both in the direct and in the indirect passive. Thus, both the direct and the indirect passive have a monoclausal structure. The clause (i) is designed to account for the presence of an Accusative marked DP in the indirect passive. Note that reassignment of the absorbed Case is a very curious property. Now notice that our analysis using the biclausal structure coupled with a general theory of Case, makes this property follow from the categorial structure of the indirect passive. Our approach can be taken to give a theoretical expression to Miyagawa’s (1989) idea.
That the impossibility of embedding the passive is not a property of the causative in general can be seen from the well-formedness of the following Japanese example. (i)Mary-wa Taroo-o Ziroo-ni home-rare-sase-ta. Mary-Top Taroo-Acc Ziroo-Dat praise-pass-cam-past ‘Mary made Taroo be praised by Ziroo.’ Marantz (1984, 273) The Japanese causative displays essentially the same behavior as the ECM construction, as noted in the literature. See section 3.4.2 below. It is therefore not surprising to find acceptable sentences like (i). The reduced causative, on the other hand, is unable to embed the passive, as is discussed in the text.
See also Zubizarreta (1985) for the impossibility of embedding the passive under the Italian causative.
Cf. Bach (1967).
One might wonder why the copula be is possible under a modal, as in (i). (i) John might be mad. We claim that a modal is adjoined to Tns and that the verb, whether the copula or not, will eventually be raised to Tns. See Chapter 4 on modals and Tns.
The suppletive relation here is defined on the basis of the infinitival form.
Suppletion may give rise to the situation in which the participial form of the be-type copula comes from the have-type copula. Kayne (1993) cites such instances and claims that direction of incoipuration is responsible for the switch. Phrased in terms of our analysis of the perfect construction, the participle takes the form of have if the lower be in (i) is incorporated into Asp, just as have is obtained if Asp is incorporated into be. (i)… [vp BE [AspP Asp [Agrp Agr BE]]] This type of suppletion is actually a variant of the second strategy for avoiding (3.105) to be discussed right below.
This binding fact poses a potential problem for the Minimalist approach. assumed here, which allows only interface conditions to apply at LF and at PF. Under this theory, the Binding Conditions are supposed to apply to the output of LF (Chomsky 1993). Since the embedded object in the reduced causative is raised to Spec of the matrix Agr-oP under our analysis, the dative subject does not c-command the raised object at LF. One possible way out is to assume that only part of the embedded object which has to do with Case checking is raised at LF, leaving the anaphoric element behind. The Binding Conditions apply to the latter. As pointed out by Jim McCloskey (personal communication), this problem is analogous to that posed by the there construction, discussed by Chomsky (1995) and Dikken (1995), where the putative LF raising of the indefinite associate should make binding in (i) possible. (i) *There seem to each other V. to have been many linguists given good job offers] Modifying Chomsky’s (1995) proposal on feature movement, we may say that only the Case and (D-features raise, leaving behind the feature having to do with binding. It should be noted at the same time that the embedded object in the reduced causative can be bound by the matrix subject, as shown in (ii). (ii) a. (?)Mariai ha fatto accusare se stessai a Giovanni. (Italian) Maria has made accuse herself to Giovanni ‘Maria had Giovanni accuse her.’ Burzio (1986, 309) b. Beni Hasan-a kendim-ii yika-t-ti-m. (Turkish) I Hasan-Dat myself-Acc wash-cans-past-1 sg ‘I made Hasan wash me.’ Aissen (1974b, 11 ) In these cases, we are led to assume that the anaphoric element is raised together with the rest of the embedded object. We will leave to future research how to accommodate these two options.
This hypothesis essentially expresses the research program of Hale and Keyser (1993), which attempts to derive thematic interpretation from a phrase structural basis. Note that under this approach, the fact that the external argument in Spec of VP allows a variety of readings depending on the choice of verbs should be due to different structural shapes that VP takes.
Baker (1988a) notes that VP fronting brings along the by-phrase, as in (i). (i)… and examined by a psychiatrist he was! The fronted constituent should be AgrP, if V is raised to the participial Agr in overt syntax. This fronting fact is consistent with our hypothesis that the by-phrase occupies Spec of VP.
We might end up saying, however, that the specifier position which hosts the by-phrase is located on the right-hand side of VP, because of the following contrast pointed out by Chris Collins (personal communication) (i) a. The house was painted red by John. b. ??The house was painted by John red. See also Baker ( 1988a, 338 – 339 ).
It is possible, however, that the null external argument is analogous to the null object in Italian discussed in Rizzi (1986b), which also fails to be identified by rich agreement. If this line of analysis turns out to be tenable, we can say that the null external argument is pro, accompanied by a null counterpart of by.
Even though the contrast between the indirect passive and the direct passive with respect to binding has been noted since Kuno (1973) and McCawley (1972), the binding facts in the niyotte direct passive have rarely been discussed in the literature. See Terada (1990), Watanabe (1993a), andTakezawa (1994).
Our treatment of the implicit external argument in this section is incompatible with the idea that theta interpretation involves feature checking. But see Saito and Hoshi (1994, note 12) for an argument that theta role assignment is feature checking. If it turns out that theta interpretation involves feature checking, we have to adopt the alternative treatment of the implicit external argument as pro, since the derivation would crash when the external argument is not represented.
This discussion is orthogonal to the A/A-bar distinction. See Abe (1993) and Lasnik (1993) for aguments against the significance of the A/A-bar distinction for binding.
This accounts for one direction of Burzio’s generalization: Accusative Case is assigned if a theta role is assigned to the subject position (8subj –> A). We turn to the other direction below.
Cf. also Marantz (1984).
Thanks to an anonymous reviewer for suggesting this line of analysis. 10 it We also have to block examples such as (i). (i) *John was received of the letter. The unacceptability of (i) should be attributed to the impossibility of using of as an identity function for an internal argument. This is a peculiarity of English, as we will see in section 3.4.1.4. Note also that essentially the same interpretive procedure should apply when the external argument is missing as in (ii), because (3.12 lb) remains anomalous even if the by-phrase is omitted. (ii) The letter was received.
There are exceptions noted in the literature. We will come back to this problem in the next subsection. Zaenen (1993) observes that the Dutch impersonal passive is sensitive to the atelicity of the construction and the controllability of the event. As long as these factors enter into determination of the external argument, the text discussion holds.
Imbabura Quechua (Jake 1983) and Malayalam (Alsina and Joshi 1991) also display the same restriction on the oblique marking of the intransitive verbs. Authier and Reed (1991) and Gibson and Raposo (1986) observe that the same contrast holds in some French dialects.
Although Radford (1977) does not use the term unaccusative or ergative, he mentions that the verbs which take essere ‘be’ in the perfect resist dative marking, attributing the observation to G. Lepschy. The use of essere is an indication of unaccusativity in Italian, as we have seen above. See Perlmutter (1978) and Burzio (1986).
This accounts for the remaining direction of Burzio’s generalization: theta role is assigned to the subject position if Accusative Case is assigned (A –> Osubj).
There is one remaining question to be settled, however. Simply saying that there is an oblique marker which can be placed on the external argument is not sufficient to account for the fact that the passive and the reduced causative use different oblique markings on the external argument in some languages. Although it is true that the faire par construction (Burzio 1986 for Italian, Kayne 1975 for French) uses the same oblique marking as the passive and displays the same Case properties as the reduced causative, it also shows properties which are not found in the reduced causative. See the references cited as well as Guasti (1992) and Zubizarreta (1985, 1987 ). Thus, in Italian, the reduced causative uses the dative marker, while the passive uses the ch -phrase. (i) a. Mario fu invitato da Giovanni. Mario was invited by Giovanni b. Maria ha fatto riparare la macchina a Giovanni. Maria has made repair the car to Giovanni We suggest that this difference is a matter of selection. Note that the participial verb is used in the passive whereas the infinitival form is used in the reduced causative. This is a difference in the agreement feature of the verb which is to be checked against the head of the embedded AgrP complement. Since the oblique external argument has to be combined with a projection of the verb, we can treat the difference in the oblique marking as a matter of selection. See section 3.4.2 for the status of selection in the Minimalist program. A piece of evidence that this line of reasoning is on the right track is provided by Chamorro, where the passive morphology is sensitive to the nature of the external argument. According to Gibson (1980, 32–34), the prefix ma-is used when the external argument is plural or unspecified, and the infix -in-is used otherwise. Since the external argument is an oblique, this cannot be agreement, but should be selection. See also section 4.4.2.3 where the Chamorro wh-agreement involving the infix -in-is compared with the passive morphology.
One might wonder how the reflexive/reciprocal morpholgy can be related to the passive morphology, given our hypothesis that the passive construction involves the copula and a participial verb. Our suggestion is that the reflexive/reciprocal arises from an A-chain having two theta positions. That is, the reflexive/reciprocal constructions also involve Accusative Case absorption, as in the passive. The difference is that Spec of VP is occupied by a trace instead of by an oblique phrase. Cf. Boskovib (1994) for general discussion on the possibility of movement into theta positions. Movement into theta positions can be better accommodated by the Agr-less Case theory of Chomsky (1995), but it goes beyond the scope of this work to pursue this possibility further.
Maling (1995) also reaches the same conclusion as we do.
What adds more confusion is the fact that some impersonal constructions display a 1AEX-like effect. Jaeggli (1986b) notes that the Spanish construction is impossible in the passive and the unaccusative predicates, though not in the subjectto-subject raising.
According to Ozkaragöz (1986), the double passive is restricted to the aorist tense.
The Turkish passive is marked by -In after vowels and /1/, and by -11 otherwise. 112 The impersonal construction discussed in this section should not be confused with the impersonal passive of the kind found in German and Dutch. The latter involves Case absorption, with the phonologically invisible internal argument undergoing LF movement to the expletive. See section 3.2.4.1 above.
According to Baker (1988a) and Bittner and Hale (1994), the oblique marking in the antipassive can appear on the ECM subject. This suggests that the oblique marking in the antipassive does not have any thematic restriction, except that it does not appear on the external argument of the verb to which the antipassive morphology is attached. The passive oblique and the antipassive oblique are indeed complementary.
Chamorro has a further complication, however, because the same oblique marking is used both for the passive and for the antipassive. The question is why the passive chooses to put the oblique marking on the external argument, while the antipassive does not. We do not have a real answer. Our tentative suggestion is that this is a matter of selection between the verb with the morphological features appropriate for the passive construction and the oblique marker. If selected, the oblique marker appears on the external argument, and we get the passive. If not, it cannot appear on the external argument. See also the note at the end of section 3.4.1.2. We will address the status of selection in the next section.
Given the similarity between the indirect passive and the causative, it is possible that the indirect passive also allows parametrization of the kind found in the causative, to which we turn right below. We do not know the facts yet.
Imbabura Quechua has both the ECM type of causative and the reduced causative.
As Collins (1994, 1995) notes, Integration is an extension of Chomsky’s (1994, 7) proposal to the effect that the derivation does not converge unless all indices in the numeration are zero.
Cf. Grimshaw (1981) and Chomsky (1986a). In the text, we essentially follow the presentation in the latter work. See also Chomsky and Lasnik (1993) for an important discussion.
This proposal does not extend to unaccusative predicates. At first sight, Pesetsky’s original proposal seems to extend to unaccusative verbs like age and seem, which do not take a DP object. Since they do not have Accusative Case, they do not take a DP object. There are run-of-the-mill unaccusative verbs like arrive, however, that take a DP object. There is no way of teasing apart these two classes of verbs except by directly mentioning c-selection.
Cf. Grimshaw (1991) and Fukui (1995) for relevant discussions.
Choice between TP and VP, on the other hand, may be reducible to s-selection of events and actions, as suggested in Rochette (1988) and Grimshaw (1991).
The literature on this topic is vast, including Kuroda (1965, 1978), Harada (1973), Shibatani (1973, 1976), Kuno (1973), Inoue (1976), Saito (1982), Teramura (1982), Kitagawa (1986), Takezawa (1987), Miyagawa (1989), Terada (1990), Dubinsky (1994), and Harley (1995), among many others.
The subject of the embedded intransitive clause can also get Dative. We will return to this point below.
Poser (1983) and Saito (1982, 1985) claim that the double-o constraint should be reduced to the statement that a verb can have at most one Accusative Case feature in Japanese. Given the biclausal nature of the causative construction, an additional constraint is needed to the effect that a complex head created through incorporation cannot exceed a simple head in the number of Case features that it has, as Baker (1988a) notes. Since incorporation itself does not play a significant role in our account, these maneuvers are unavailable to us. We will see below that our framework provides a simpler account.
It should be noted, however, that Dative marking on the intransitive subject does not always force the permissive or ‘let’ reading, as is sometimes claimed. See Inoue (1976, 70) for a critical discussion. See also Kuroda (1992, 290), Takezawa (1987), and Terada (1990, 234–237). The text discussion exclusively deals with the ordinary causative reading roughly corresponding to ‘make’.
The examples in (3.144) and in (3.145) are taken from Harada (1973, 130) and Teramura (1982, 297), respectively, with slight modifications.
Teramura (1982, 298) notes that the verbs of movement which mark the path argument with Accusative are not comfortable with the Dative marking, as in (i). (i) ?Kodomo-ni tsuna-o watar-ase-ru. child-Dat rope-Acc cross-caus-pres ‘to make children cross the rope’ These verbs are shown to be unaccusative by Miyagawa (1989).
There is one more assumption that we have to make. The infinitival Tns node in the TP causative complement cannot possess a Null Case feature, since there is no Agr to take care of its absorption. This is a marked property of the Japanese causative.
See also Travis (199 lb) for a similar approach.
This is a nontrivial assumption which needs to be justified, though. Under the VP-internal Accusative hypothesis, internal arguments of the verb appear within the lower VP and the external argument is placed in Spec of the higher VP in (3.148b). Thus, the traditional VP is split up in two domains separated by a functional projection. One might invoke the Aspect head proposed by Travis (1991b) for this particular purpose. Strictly speaking, this head should not be confused with the Aspect head which we have proposed above. Travis’ Aspect head is responsible for the VP-internal aspect.
Another possibility is to get rid of Agr as in Chomsky (1995). Under Chomsky’s (1995) proposal, transitive verbs have the structure (i). Thus, the higher verb checks Accusative Case. If the original subject position is higher than the position for Accusative Case checking, the relevant structure will be: If the Accusative Case checking position is higher than the original subject, the relevant configuraiton is: Ura (1996) argues that UG allows both (ii) and (iii). Either way, the head of the higher VP, Vb, must undergo follow-up checking with a higher functional head, Tns in ordinary full transitive clauses. Thus, irrespective of the relative height of the original subject and the derived object, our Layered Case Checking Theory carries over.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 1996 Springer Science+Business Media Dordrecht
About this chapter
Cite this chapter
Watanabe, A. (1996). V → AGR-O → TNS/ASP. In: Case Absorption and WH-Agreement. Studies in Natural Language and Linguistic Theory, vol 37. Springer, Dordrecht. https://doi.org/10.1007/978-94-015-8615-3_3
Download citation
DOI: https://doi.org/10.1007/978-94-015-8615-3_3
Publisher Name: Springer, Dordrecht
Print ISBN: 978-90-481-4744-1
Online ISBN: 978-94-015-8615-3
eBook Packages: Springer Book Archive