Two German Impersonal Passives and Expletive pro

Abstract In the present article I argue that the null subject observed in German impersonal passives is neither a null expletive nor in canonical subject position (e.g. AgrSP). We will link the possibility of null subjects in German constructions involving the participle to the possibility of nominative assignment into the VP. The null subject is then analysed as a syntactically projected null cognate object that is assigned nominative case in situ. Key words:null expletives, nominative-assignment, German impersonal passives. Resum. Dues passives impersonals de l’alemany i l’expletiu pro En aquest article argumento que el subjecte buit de les passives impersonals de l’alemany no és ni un expletiu buit ni a la posició canònica del subjecte (e.g. SconcS). Relaciono el fet que els subjectes buits siguin possibles a les construccions de participi de l’alemany amb la possibilitat d’assignació de cas nominatiu dins del SV. El subjecte buit s’analitza doncs com un objecte cognat buit projectat sintàcticament que rep cas nominatiu in situ. Paraules clau:expletius buits, assignació de nominatiu, passives impersonals de l’alemany. CatWPL 7, 199947-57 Two German Impersonal Passives and Expletive pro Patricia Cabredo Hofherr Université de Paris 7, URA 1028 2, Place Jussieu. 75251 Paris Cedex 05. France [email protected] Received: December 13th 1998 Accepted: March 17th 1999 Table of Contents CatWPL 7 047-057 13/6/00 12:24 Página 47 1. Introduction 2. The Semi-pro-drop Analysis 3. Seinand werden-passive 4. The Analysis 5. Conclusion References 1. Introduction The German impersonal passive construction with werden poses a well-known puzzle of German syntax: even though German is not a pro-drop language (see (1)), the subject position of the impersonal werden-passive must remain empty (see (2)): (1) Gestern kam *(er) zu spät. yesterday came he too late ‘Yesterday he came too late.’ (2) Gestern wurde (*es) lange diskutiert. yesterday werden 3sg expl long discussed ‘Yesterday it was discussed/ the discussion went on until late.’ When the expletive es is inserted in subject position of a werden-passive it receives an argumental reading as the 3sg neuter pronoun where possible (see (3)), resulting in ungrammaticality when the argumental reading is impossible as in (4). (3) Gestern wurde es früh gegessen. yesterday werden 3sg it early eaten ‘Yesterday it (e.g. the food) was eaten early.’ (4) *Gestern wurde es früh geschlafen. yesterday werden 3sg it early slept ‘Yesterday it (=sth) was slept early.’ 2. The Semi-pro-drop Analysis The standard analysis for the empty subject position in German impersonal passives within the generative framework proposes that German is a semi-pro-drop language (e.g Sternefeld (1985), Safir (1985), Grewendorf (1989)). A full pro-drop language such as Spanish licenses null referential pronouns (see (5a)) while a restricted pro-drop language such as German is taken to allow only the equivalent of (b): a phonetically empty expletive pronoun proexpl. (5) a. (Sp) pro hablo speak 1sg ‘I speak.’ b. proexpl es evidente que quieren invitarnos. is evident that (they) want invite-us ‘It is evident that they want to invite us.’ An empty expletive explains why the subject position of the impersonal passive can remain empty: it is filled by the proexpl. The fact that it must be empty is then derived by the Avoid Pronoun Principle (see Safir (1985)). 48 CatWPL 7, 1999 Patricia Cabredo Hofherr CatWPL 7 047-057 13/6/00 12:24 Página 48 (6) Avoid Pronoun Principle (APP) (Chomsky (1981)): Wherever an overt pronoun can be avoided it must be avoided. 2.1. The Distribution of Expletives in German The semi-pro-drop analysis assimilates the null subject in German impersonal passives to expletives. This raises a number of problems. In German the impersonal use of actives verbs (see (7)) and weather verbs (see (8)) do not allow expletive drop: (7) Vorhin hat *(es) am Fenster geklopft. just now has expl at the window knocked ‘Just now there was a knock at the window.’ (8) Heute schneit *(es). today snows expl ‘Today it is snowing.’ Under the plausible assumption that weather-arguments are not expletives but rather quasi-arguments, it has been proposed that the German proexpl corresponds to the true expletive while es is a semi-argument. An analysis in terms of proexpl and the APP cannot be extended to another case of null subject positions that arises in German. In some cases of extraposition es can be left out (see (9a)); that this is not a property of extraposition in general as shown in (9b)): (9) a. …dass (es) klar war, dass es so kommt. … that it clear was that it so comes ‘…that it was clear that that would happen.’ b. …dass *(es) keinen Verdacht erregte, dass… …that it no suspicion roused, that… ‘… that it aroused no suspicion that …’ The APP would lead us to expect that where es can be dropped it is impossible to insert it. This expectation is not borne out: in extraposition constructions es is always possible in subject position. Furthermore, the proexpl postulated for the impersonal passive cannot be taken to fill the empty position in (9) since languages that have two overt expletives corresponding to a true expletive and a semi-argument only admit the argumental variant in extrapositions. (10) (Dutch) … dat het/*er mij irriteert dat … (Danish) …at det/*der irriterer mig at … …that it/*there annoys me that … (examples taken from Vikner (1995:235)) Two German Impersonal Passives and Expletive pro CatWPL 7, 1999 49 CatWPL 7 047-057 13/6/00 12:24 Página 49 The element proexpl cannot appear in this configuration. This is equally the case for other impersonal constructions like the impersonal middle where the expletive es subject is obligatory: (11) Hier lebt *(es) sich gut. here lives expl refl well ‘Here it lives well / here people live well.’ (impersonal middle) Under a proexpl-analysis of the impersonal passive the argumental status of the middle subject has to differ from that of impersonal passives. Except for the difference with respect to expletive surface subjects there is no evidence for such a difference of theta-roles. 2.2. Conclusion The semi-pro-drop analysis for German raises a number of problems briefly reviewed above. We have seen that the distribution of expletives in German depends on the syntactic construction involved, and that the distribution of proexpl is limited to a particular construction, namely the impersonal passive. In what follows I will therefore concentrate on two construcions where the lexical expletive is systematically excluded: the seinand the werden-passive. I will argue that the lack of an overt subject in the German impersonal passives is not to be found in a global property such as «semipro-drop» but rather in the properties of the past participle and its interactions with the syntactic environment it appears in. In particular I propose that the null subject in the passive corresponds to the null object that is present with active intransitive verbs. 3. Seinand werden-passive 3.1. The Sein-passive The German sein-passive is a stative passive formed with the auxiliary sein, ‘to be’, and the past participle. Like the werden-passive it allows a personal (12) and an impersonal (13) variant. (12) Der Garten ist gegossen. The garden is watered (=stative/resultative reading) (13) Er sah, dass eingekauft war. he saw that shopped was ‘He saw that the shopping had been done.’ Like the werden-passive the impersonal sein-passive cannot take an expletive subject (see e.g. Rapp (1997: 214)): 50 CatWPL 7, 1999 Patricia Cabredo Hofherr CatWPL 7 047-057 13/6/00 12:24 Página 50 (14) a. Er sah, dass schon serviert war. he saw that already served was ‘He saw that the serving had already been done.’ b. Er sah, dass es schon serviert war. he saw that it already served was ‘He saw that it (the food) had already been served.’ 3.2. Interpretive Differences The two passives have clearly different interpretive properties. The werden-passive shares the aspectual properties of the active verb while the sein-passive is always stative picking out the resultant state of the action described by the verb. The werdenpassive can combine with all the modifiers that can modify the active verb (see (15)); the sein-passive only allows modification that affects the resultant state (see (16)). (15) Der Brief wird langsam/mit roter Tinte geschrieben. the letter werden3sg slowly/with red ink written ‘The letter is being written slowly/with red ink (process).’ (16) Der Brief is *langsam/mit roter Tinte geschrieben. the letter sein3sg *slowly/with red ink written ‘The letter is written slowly/with red ink (result).’ The werden-passive implies disjoint reference between the surface subject and the implied agent, the sein-passive allows both reflexive and disjoint interpretation (see(17)). (17) Hans ist gewaschen. ‘Hans is washed. ‘ (resultative) ok: Hans/somebody else washed Hans. 4. The Analysis 4.1. The Sein-passive Levin & Rappaport (1986) propose that the English adjectival passive involves a lexical category changing operation on the past participle: (18) V[Part]→[V [Part]]ALevin & Rappaport (1986: 646) For German this analysis is too restricted. Kratzer (1996) shows that the German adjectivised past participle can be phrasal, in that it can retain the dative argument of the underlying verb. (19) Das Land war ihm abgeschmeichelt. the land was him (dative) off-flattered ‘The land was obtained from him by flattery’ (Kratzer’s example) Two German Impersonal Passives and Expletive pro CatWPL 7, 1999 51 CatWPL 7 047-057 13/6/00 12:24 Página 51 Following Kratzer I will therefore assume that the adjectival past participle in German is derived by affixation of an adjectival affix. This affixation may be lexical (20) or phrasal (21): This analysis presupposes that the external argument of the verb is introduced outside the maximal projection minimally containing the object. I will assume with Kratzer that the external argument is the argument of a specialised VOICE head that may also introduce accusative case. Under these assumptions the fact that the (acc-assigning) present participle cannot be adjectivised (see (21)) shows that the adjectival affix cannot take a VoiceP complement. (21) *Der Mann ist seine Kinder liebend. the man is his children loving Adjectivisation of the participle therefore necessarily prevents the projection of the VoiceP. Since the VoiceP is absent in adjectival participles, the external argument is syntactically absent — see Kratzer (1996) (cf. reflexive interpretation of the implicit argument). The overall structure of the syntactic as well as the lexical variant of the German sein-passive is a copula construction parallel to the combination of sein plus an underived adjective as in (22b) (see e.g. Höhle (1978)). (22) a. Das Haus ist gestrichen. The house is painted. (sein-passive) b. Das Haus ist blau. The house is blue. (copula sein + adjective) 4.2. The Werden-passive For the werden-passive as for the sein-passive the participle first projects its internal argument. I propose to analyse werden as a voice-auxiliary: it fills the Voice head, allowing the projection of the VoiceP: 52 CatWPL 7, 1999 Patricia Cabredo Hofherr (20) A (21) A AVAVP (lexical adj. part) (phrasal adj. part.) (23) VoiceP Voice’ Voice° VP werden object V’ V CatWPL 7 047-057 13/6/00 12:24 Página 52 As the verb werden cannot assign accusative (see (24)), only one argument can be realised. (24) Er wird der neue Gesundheitsminister. He (nom) becomes the new health secretary (nom) The internal argument is realised since it is immediately merged within the VP; it is therefore already present when the second, external, argument has to be legitimised. As the impersonal variants of the seinand werden-passives rely on the same auxiliaries combined with the past participle as their personal counterparts, I will assume that personal and impersonal passives have the same syntactic structure. This forces us to assume that unergative intransitive verbs in German have a syntactically realised null cognate object. Adapting an analysis by Dobrovie-Sorin (1994) for Romanian passives I will assume that the empty object has to stay in situ since movement out of the VP is incompatible with its semantic content (cf. the difference in interpretation between VP-internal and VP-external bare plurals in German, Diesing (1992)). If the null cognate object cannot move out of the VP, how is it licensed? In what follows I will argue that it is licensed in situ through the assignment of [nom] case. The possibility of having an impersonal passive without an overt subject in German is then linked to the possibility of assigning [nom] to the object position in the werden-passive. That in German [nom] can be assigned into the VP is supported by the fact that the passive of double-object constructions has the subject of the sentence in the same position as the [acc] object in the active (see e.g. den Besten (1985), Broekhuis (1992)). (25) a. dass Hans einem Jungen dieses Buch gegeben hat. that Hans to a boy (dative) this book(acc) given has ‘that Hans gave this book to a boy.’ b. dass einem Jungen dieses Buch gegeben wurde. that to a boy (dative) this book(nom) given was ‘that this book was given to a boy.’ In Dutch, another Germanic OV-language, there is also some evidence that the Dutch cognate of the German werden-passive does not involve NP-movement. In active double object constructions in Dutch the indirect object can be realised by a bare NP or by a PP introduced by aan. If the indirect object is realised as an NP it has to precede the direct object: (26) a. dat de jongen (aan) het meisje het cadeau gaf. that the boy (to) the girl the present gave b. dat de jongen het cadeau *(aan) het meisje gaf. that the boy the present (to) the girl gave ‘that the boy gave the present to the girl.’ Two German Impersonal Passives and Expletive pro CatWPL 7, 1999 53 CatWPL 7 047-057 13/6/00 12:24 Página 53 Now consider the distribution of aan in the double object passive. The data in (27) suggest that (27b) is not derived from (26a) by NP-movement but related to the acc-dative order in (26b). (27) a. dat (aan) het meisje het cadeau werd gegeven. that (to) the girl the present was given b. dat het cadeau *(aan) het meisje werd gegeven. that the present (to) the girl was given ‘that a present was given to the girl.’ 4.3. No Expletive Subject In both the phrasal sein-passive and the werden-passive the underlying object initially occupies the object position (spec VP). In the impersonal variants the null object has to be licensed in situ by transmission of [nom] case. Since the null object needs case the lexical expletive in subject position is excluded for case reasons. This analysis implies that the es in preverbal position in presentative V2 configurations is not inserted in subject position but directly in spec CP (as suggested for thad, the Icelandic equivalent of es in Platzack & Holmberg (1995)): (30) [CP Es [C’ wurde [IP [VP null object [V’ getanzt.]]]] expl werden 3sg danced ‘It was danced/ There was dancing.’ The similarity between impersonal phrasal sein-passives and werden-passives is obscured by the fact that lexical sein-passives admit the impersonal use like underived adjectives: this use always takes an es subject. (31) a. Bei dir ist (es) immer so aufgeräumt. at your place is expl always so tidied up ‘At your place it is always so tidy.’ (lexical sein-passive) b. Bei dir ist *(es) immer so ordentlich. at your place is expl always so tidy ‘At your place it is always so tidy.’ (underived adjective) 54 CatWPL 7, 1999 Patricia Cabredo Hofherr ((28) VP (29) VoiceP sein AP Voice’ A VP werden VP object V object V (sein-passive) (werden-passive) CatWPL 7 047-057 13/6/00 12:24 Página 54 In contrast with underived adjectives, however, sein-passives also allow the expletive-less (i.e. phrasal) construction. 4.4. The Interaction of Werden with the Perfect The analysis of werden as a Voice auxiliary is supported by other evidence showing that the link between passive werden and the lexical verb is tighter than that between a verb and its perfect auxiliaries. Several languages show an interaction of perfect and argument structure (see van den Wyngaerd (1996) citing Postma (1994)). One example is provided by German modals: present tense modals have a root and an epistemic reading while perfect modals only have the root reading: (32) Er hat lachen müssen. he has laugh must (infinitive) ok: ‘He had to laugh.’ (root) *: ‘It is/was probable that he laugh/laughed.’ (epistemic) In German the argument changing effect of the perfect interacts with the argument-reducing passive werden: in the perfect the participle of werden appears without the ge-prefix typical of the past participle: (33) *Heute ist lange gearbeitet geworden. Heute ist lange gearbeitet worden. today is long worked werden past part ‘Today it was worked for a long time.’ The past participles of the sein-passive, gewesen (see (34)),and of the perfect auxiliary haben, ‘to have’, keep their ge-prefix in the periphrastic perfect: (34) a. Das Haus ist angestrichen. The house is(copula) painted. b. Das Haus ist angestrichen gewesen. The house is(copula) painted been (35) a. Er hat angerufen. he has rung b. Er hat angerufen gehabt. he has rung had This difference between passive werden and the other functional verbs follows from the analysis proposed above if we assume the structure for the past participle suggested in van den Wyngaerd (1996). He proposes that the ge-prefix of a participle is generated in a separate position below the verb, attaching to the verb root by head movement: Two German Impersonal Passives and Expletive pro CatWPL 7, 1999 55 CatWPL 7 047-057 13/6/00 12:24 Página 55