GGS 2007 Konstanz, 18.05.07

On the realization of external arguments in nominalizations Artemis Alexiadou & Florian Schäfer University of Stuttgart artemis/[email protected]

1.

Introduction

In this paper we investigate the realization of external arguments in nominalizations. We want to address the question how syntactic structure in nominalization and encyclopedic knowledge of roots interact in realizing and projecting external arguments in this domain. We focus on nominalizations derived from verbs undergoing the causative-anticausative alternation in mainly German, which we will analyze following the general tenets of syntactic approaches to nominalization. 1 •

We take the causative-anticausative alternation as our test-case, as a) a consensus exists concerning the encyclopedic classification of verb meanings, which predicts which verbs alternate and which do not (Haspelmath 1993, Levin & Rappaport-Hovav 1995 and others). b) external arguments of verbs undergoing the alternation can be associated with two types of thematic roles (Agent, Causer) c) these thematic roles can be realised as DPs or PPs, whereby different prepositions are used for the Agent and the Causer argument (in German von vs. durch respectively). d) the morphology some verbs bear in German helps to disambiguate them (they are interpreted as anticausative when they surface with the reflexive pronoun sich). e) there is a certain consensus concerning the structural analysis of the verbs undergoing the alternation.

Questions: 1. In the verbal domain, external argument realization is subject to formal requirements. Is argument realization in nominalization subject to the same principles as in the case of verbs? 2. Can we identify systematic patterns?

1

In our discussion we abstract away from the differences between complex event and result nominals, as discussed in Grimshaw (1990). For our purposes, all the nominals discussed here are complex event or argument supporting ones in Grimshaw's terms.

2. 2.1

The (anti-)causative alternation The verbal pattern

(1)

a. b. c.

John broke the window The storm broke the window The window broke (from the storm)

TABLE 1. DISTRIBUTION OF ARGUMENTS IN GERMAN Causatives Passives Yes (Subject) Yes (von-phrase) Agents Yes (Subject) Yes (durch-phrase/ Causers *von-phrase) (2)

a. b.

(3)

a. b.

•

(4)

Anticausatives No Yes (durch-phrase/ *von phrase)

Hans / der Erdstoss zerbrach die Vase Hans / the earth tremor broke the vase Hans / der Wind öffnete die Tür Hans / the air opened the door Die Vase wurde von Peter / durch den Erdstoss zerbrochen The vase was by Peter / through the earth tremor broken Die Tür wurde von Peter / durch einen Windstoss geöffnet The door was by Peter / through a gust-of-wind opened

There are two morphologically distinct types of anticausatives in German (± reflexive morphology), they behave alike with respect to argument realization. a. b.

2.2

Agent Causer Anticausative

Die Vase zerbrach (durch ein Erdbeben) The vase broke through an earthquake Die Tür öffnete sich (durch einen Windstoß) The door opened REFL through a blast-of-wind

A syntactic analysis of the alternation

•

Following Marantz (1984), Kratzer (1994): external arguments are not arguments of verbs but of a Voice head.

•

The structure of (anti-)causative verbs: (Alexiadou et. al. 2006) 2

(5) •

[ (Voice) [ vCAUS(e) [ Root + Theme ]]]

Agentivity and causation are syntactically represented in terms of distinct functional heads. CAUS is taken to introduce a causal relation between a causing event (the implicit argument of CAUS) and the resultant state denoted by the Root + Theme complex. Voice introduces the external argument.

2

Note that CAUS could also simply be seen as an eventive v of the type proposed in Marantz (2005). In this case the causative semantics would not be directly encoded on any verbal head but would result from the combination of an activity v and its stative complement (see Ramchand 2006 for related ideas).

2

•

Causatives and Anticausatives have a CAUS head but differ in that only the former have Voice. The existence of CAUS in Anticausatives can be detected by the crosslinguistic licensing of Causer-PPs (German: durch; English: from; Greek: apo; …).

• •

Roots fall into different classes depending on their Encyclopedic semantics (cf. also Bhatt & Embick in progress, Harley & Noyer 2000): 1. 2. 3. 4.

√agentive √internally caused √externally caused √cause unspecified

(murder, assassinate) (blossom, wilt) 3 (destroy, kill) (break, open)

•

All of the above roots combine with CAUS.

•

‘Internal vs. external causation’ categorization of the root influences the combinations of roots with particular types of Voice heads.

Agentive roots:

The bringing about of the event requires the presence of an Agent. They combine with Voice and are necessarily transitive.

Internally caused:

The cause of the change of state event is linked to properties inherent to the argument undergoing change and cannot be directly caused by an external argument. They do not alternate and are necessarily intransitive.

Externally caused:

The change of state is brought about by an external cause. They combine with Voice and are necessarily transitive.

Cause unspecified:

No specification of internal vs. external cause. They alternate.

•

How can we explain the presence of sich, i.e. the fact that in German anticausative verbs split into two groups, those that take sich and those that do not?

•

Haspelmath (1993) established a correlation between the presence/absence of extra morphology in anticausatives and the conceptualization of the change of state event expressed by the verb. Verbs (roots) are ordered on a ‘spontaneity scale’. The less likely it is for a change of state event to happen spontaneously the more likely it is that extra morphology shows up. (To a certain extent this is idiosyncratic).

•

Schäfer (in preparation) argues that roots that necessarily come with extra morphology have the requirement to appear in the presence of Voice, even if they express an anticausative event. In this case, a special kind of Voice is involved (expletive Voice with no semantic content). Non-referential, non-bound sich is located in the Specifier of this Voice. Other languages (e.g. Greek) use non-active morphology without semantic impact for this function.

3

For us unergative predicates are not causatives, and hence cannot be classified as internally caused, contra Levin & Rappaport (1995) and in line with Marantz (1997).

3

(6)

Interpretation:

Syntax:

Spell-out:

anticausative: anticausative: causative:

[ V [ RootA + Theme ]] [DPexpl. [Voice{D, ∅} [ V [ RootB + Theme ]]]] [DP [Voice{D, Agent} [ V [ RootA/B + Theme ]]]]

∅ (4a) sich (4b) transitive

B

•

How do the nominalizations of these different root classes behave?

•

In principle, there are two possible realizations for the external argument in the DP: possessor or PP.

•

But, German shows restrictions as to the type of elements that can appear in the possessor position. Only proper names are allowed. Therefore we restrict ourselves to PPs.

•

Can we find the same PPs in nominalization?

•

What about sich?

3. 3.1

Nominalization patterns Background: English 4

(7)

a. b.

John's destroying the manuscript John's/the destruction of the manuscript

gerund -ation nominal

Observation 1: the realisation of the external argument is optional for -ation, even though on the basis of the root meaning, one would necessary expect the presence of an external argument (externally caused root). The same holds for agentive roots. (8)

*The manuscript destroyed

•

Why is the root's behavior different in the nominal domain?

•

Kratzer (1994), Marantz (1997 and subsequent work): 'external' arguments are never assigned by the lexical entry, but by Voice. -(at)ion nominals lack Voice and therefore they never have an 'external' argument. 5

•

When the nominalization structure includes Voice, the external argument is realized obligatorily.

•

When the nominalization lacks Voice, then the external argument can be realized as a possessor. In the case of destruction, the possessor can be interpreted as an agent, based on our encyclopedic knowledge about destroy.

4

Note that here, we take the ing of and the -ation nominalization form to be similar, following Chomsky (1970). It follows from this that they cannot assign accusative Case to their internal argument, as opposed to the gerund, which contains Voice, has an external argument and can assign accusative.

5

4

•

Note that in the absence of a possessor the interpretation of these nominalizations is passive (Alexiadou 2001, Borer 2001). Observation 2: nominalizations of alternating verbs seem to behave like their verbal counterparts, 6 with one major difference (ambiguity of the intransitive variant): (9)

a. b. c. d.

John's breaking of the vase the breaking of the vase (by John/by the wind) John's accumulation of wealth the accumulation of wealth (by John)

•

(9b-d) are ambiguous between two interpretations: a passive and a spontaneous, anticausative, reading. 7

•

In the verbal domain the two interpretations are morphologically encoded (i.e. the passive is clearly distinct from the anticausative structure).

3.2 • (10)

German Three types of nominalizations to be considered (Ehrich 1991, Ehrich & Rapp 2000): a. b. c.

Das Heilen der Wunde the healing the wound-gen Das eine Wunde Heilen the one wound-acc healing Die Heilung der Wunde the healing the wound-gen

Root-en + genitive Root-en + accusative Root-ung + genitive

3.2.1 Root+en These nominalizations show verbal properties. i) they license lexical case

(11)

das dem Nachbarn helfen the the-dat neighbor help-en

ii) they can include modal verbs

(12)

das Studieren Wollen the study-inf want-en

iii) they allow passive auxiliaries and participles (13)

um [allzuhäufiges Überfahren und Verhaftet Werden] zu vermeiden in-order-to all-too-often overrun.Pass.Part and arrested.Pass.Part become to prevent ‘in order to prevent that (one) is overrun or arrested too often …’

6

Marantz (1997) and Harley & Noyer (2000) discuss the behavior of grow. Since we do not think that the two variants belong to the same verb, we do not discuss this here. See Borer (2001) for an alternative analysis: (i) a. John grows tomatoes b. Tomatoes grow c. *John's growth of tomatoes d. the growth of tomatoes e. Mary's growing of the tomatoes 7 Harley & Noyer (2000) note that if roots or even root +theme combinations can never be conceived of as being externally caused, then they never have a possessor interpreted as an external argument. (i) a. Wealth/Dust accumulated b. John accumulated wealth/#dust c. John's accumulation of wealth/#John's accumulation of dust

5

iv) they allow active, passive and anticausative realizations with or without extra morphology. The transitive pattern A: the internal argument has accusative, the external argument appears postnominal and has genitive. (14)

a. b. c.

Das grossartige eine Wunde Heilen des Medizinmannes the extraordinary a-acc wound heal-en the-gen medicine-man das häufige verrostete Türen Öffnen des Opas the often rusted doors-acc open-en the-gen grandfather das häufige die Wiesen Überschwemmen des Flusses the often the-acc meadows flood-en the-gen river

The transitive pattern B:

(15)

a. b. c.

the internal argument has accusative and there is no overt external argument. Since this is the only case nominalizations where introduction of an external argument via PP is blocked, we conclude that an external argument is covertly represented. We take this to be PRO.

Das grossartige eine Wunde Heilen (??durch / ??vom Medizinmann) the extraordinary a-acc wound healen through / by the medicine-man das häufige verrostete Türen Öffnen (??durch / ??von dem Opa) the often rusted doors-acc open-en through / by the grandfather das häufige die Wiesen Überschwemmen (??durch den Fluß) the often the-acc meadows flood-en through the river

The passive pattern: the internal argument has genitive and the agent can be introduced via a durch-phrase. The von-phrase is ungrammatical. A second genitive is ungrammatical, too. (16)

a. b. c.

das Heilen der Wunde durch den Medizinmann the heal-en the-gen wound through the medicine-man *das Heilen der Wunde vom Medizinmann the heal-en the-gen wound by-the medicine-man *das Heilen der Wunde des Medizinmannes the heal-en the-gen wound the-gen medicine-man

The anticausative pattern:

(17)

a. b.

•

the internal argument has genitive and the causer can be introduced via a durch-phrase and nothing else.

das Heilen der Wunde durch die Kamille(-ngabe) the heal-en the-gen wound through the camomile(-giving) *das Heilen der Wunde von der Kamille(-ngabe) the heal-en the-gen wound by the camomile(-giving)

Does it make sense to distinguish between the passive and the anticausative pattern?

YES 1: if -en nominals are really nominalizations of verbal structures, we would expect that this difference is real. YES 2: nominalizations of verbs that form marked anticausatives make a morphological distinction between the two patterns. 6

•

(18)

With these verbs we find both patterns (with and without sich). The two differ in systematic ways. The sich pattern (anticausative) is compatible only with causer PPs, while the non-sich pattern (passive) is compatible only with agent PPs (both introduced via durch). (Note that the German verbal passive licenses both theta roles. This pattern resembles what we found in the verbal syntax of e.g. Hebrew and Greek.) a. b. c. d.

(19)

a. b. c. d.

•

(20)

??das sich Öffnen der Türen durch Peter the REFL open-en the-gen doors through Peter das Öffnen der Türen durch Peter the open-en the-gen doors through Peter *das Öffnen der Türen von Peter *das sich Öffnen der Türen von Peter (?)das sich Öffnen der Türen durch den Wind the REFL open-en the-gen doors through the wind *das Öffnen der Türen durch den Wind the open-en the-gen doors through the wind *das sich Öffnen der Türen vom Wind *das Öffnen der Türen vom Wind

Adding the adjective spontan (spontaneous) clearly gives two different interpretations: in the absence of sich it modifies an implicit agent's action; in the presence of sich it modifies the unfolding of the anticausative event (-en nominalizations of unmarked verbs are ambiguous between the two readings): a. b.

das spontane Öffnen der Türen the spontaneous open-en the-gen doors das spontane sich Öffnen der Türen the spontaneous REFL open-en the-gen doors

TABLE 2. THE DIFFERENT CONSTRUALS OF –EN NOMINALIZATIONS Verb- Semantics int. Arg. ext. Arg. ext. Arg. Construal Type realisation semantics realisation causative acc. agent/causer gen. Transitive Voice ∅ Sich causative acc. agent/causer gen. Transitive Voice ∅ Sich

causative causative

acc. acc.

agent/causer agent/causer

PRO PRO

Infinitival Voice Infinitival Voice

∅ Sich

causative causative

gen. gen.

agent agent

(durch) (durch)

Passive Voice Passive Voice

∅ Sich

anticausative gen. anticausative gen.

causer causer

(durch) (durch)

Anticausative: no Voice Expletive Voice 8

8

It seems that some verbs that necessarily use a reflexive to form verbal anticausatives can have an anticausative reading in –en nominalizations in the absence of the reflexive (Kaufmann 2003). It seems to us that this phenomenon is restricted to a number of verbs. Even with these verbs the reflexive can be used to disambiguate the anticausative reading. (i) weil *(sich) die Krankheit / das Feuer ausbreitet because REFL the disease / the fire spreads

7

3.2.2 Root + ung •

Nominalizations of externally caused roots can appear without an external argument. Only a proper name can appear in the possessor position; the interpretation of the nominalization in (21a) is passive. The same holds for agentive roots.

(21)

a. b. d e.

• •

die Zerstörung der Stadt the destruction the-gen city Peters Zerstörung der Stadt Peter's destruction the-gen city *des Sturmes Zerstörung der Stadt the storm's destruction the-gen city *Die Stadt zerstörte (sich) the city destroyed REFL

Is the syntax also passive? -Ung nominalizations of alternating verbs: the class of verbs with marked morphology in the anticausative pattern cannot take sich in the -ung nominalization. Does this mean that an anticausative interpretation is blocked?

(22)

*die sich Öffnung der Tür the REFL open-ung the-gen door

NO 1: The nominalization licenses PPs introducing both agents and causers: (23)

a. b.

(24)

a. b.

(25)

a. b.

(26)

a b.

die Öffnung der Tür durch Peter the open-ung the-gen door through Peter *die Öffnung der Tür vom Peter the open-ung the-gen door by Peter die Öffnung der Tür durch den Wind the open-ung the-gen door through the wind *die Öffnung der Tür vom Wind the open-ung the-gen door by-the wind die Heilung der Wunde durch Peter the heal-ung the-gen wound through Peter *die Heilung der Wunde von Peter the heal-ung the-gen wound by Peter die Heilung der Wunde durch die Kamille the heal-ung the-gen wound through the camomile *die Heilung der Wunde von der Kamille the heal-ung the-gen wound by the camomile

NO 2: Addition of the adjective spontaneous leads to ambiguity with respect to the presence of an implicit agent and the unfolding of the anticausative event: (ii)

das rasche (sich) Ausbreiten der Krankheit / des Feuers the fast REFL spread-en the-gen disease / the fire

8

(27)

a. b.

die spontane Veränderung der Beitragssätze (passive) the spontaneous alter-ung the-gen rates-of-contribution die spontane Veränderung des Blutbildes (anticausative) the spontaneous alter-ung the-gen blood-count

•

Are the two interpretations syntactically encoded or not?

•

The same question arises for English -ation nominals, and both perspectives have been proposed in the literature (No: Grimshaw 1990, Alexiadou 2001, Yes: Borer 2001, van Hout & Reoper 1998).

•

Intensifying SELBST: the intensifier selbst (itself) in German has two uses; an adnominal and ad-verbal/agentive use.

•

In the adnominal use it provides a set of (contextually restricted) alternatives to the modified noun (Eckardt 2001).

(28)

Der König SELBST öffnete die Türe. the king himself opened the door {f |f maps king onto someone who might have opened the door instead} = {butler-of, servant-of, child-of, maid-of, . . .}

•

(29)

Hole (2006) argues that in the agentive use SELBST modifies Voice. It provides us with a set of (contextually given) alternatives to the way the original agent is thematically involved in the proposition. Q: Hat Paul den Kuchen SELBST gebacken? has Paul the cake himself baked ‘Did Paul bake the cake himSELF?’ A1: NEIN, jemand hat ihn FÜR ihn gebacken. no someone has it for him baked ‘NO, someone baked it FOR him.’ A2: NEIN, er hat ihn jemanden backen LASsen. no he has it someone bake let ‘NO, he HAD someone bake it.’ A3: NEIN, er hat ihn in einem LAden geKAUFT [wo er frisch gebacken WORden ist]. no he has it in a shop bought where it had just been baked ‘NO, he BOUGHT it from a SHOP [where it HAD just been baked].’

•

(30)

The licensing of agentive selbst can therefore be used as a test for determine the presence of Voice in nominalization. Wir sollten Studierende besser beraten. Das Studierende-Beraten … we should students.ACCstrong better counsel. The students.ACCstrong-counseling ‘We should counsel students better. The counseling of students …’ 9

(31)

a.

b.

(32)

a. b.

•

das SELBST-Studierende-Beraten the self-students-counseling ‘counseling students oneself’ das Studierende-SELBST-Beraten the students-self-counseling ‘counseling students oneself’ *die (SELBST-)Studierend-en-Beratung the self-student-LINKER-counseling *die Studierend-en-(SELBST-)Beratung the student-LINKER-self-counseling

The test suggests that -ung nominalizations do not contain Voice. The fact that an anticausative interpretation is possible is not surprising then. However, the passive interpretation is unexpected.

TABLE 3. THE DIFFERENT CONSTRUALS OF –UNG NOMINALIZATIONS Verb- Semantics int. Arg. ext. Arg. ext. Arg. Construal Type realisation semantics realisation causative gen. agent/causer (durch) ? no Voice ∅ sich causative gen. agent/causer (durch) ? no Voice ∅ sich •

(33)

4. • (34)

anticausative gen. anticausative gen.

causer causer

(durch) (durch)

no Voice ? no expletive Voice

(33) shows the external argument realization with an externally caused verb and an internally caused verb. Under a passive reading, agents and causers are licensed, while under an anticausative reading only causers are licensed: a.

Die Zerstörung der Stadt durch Caesar/ durch das Erdbeben The destroy-ung the-gen city through Caesar/the earthquake

b.

Die Verarmung der Bevölkerung durch die jahrelange Dürre/ The impoverish-ung the-gen people through the years-long drought/ *durch die Bundeskanzlerin through the Chancellor

External Argument implication in the absence of Voice The structures for different nominalization types: a. b.

[DP [ -en [(Voice) [VP ...]]]] [DP [ -ung [VP ...]]]

The question we wanted to address is how the real-world, i.e. encyclopedic, knowledge interacts with formal properties underlying well-formedness. •

-ung/-ation nominals suggest that the transitivity requirement for externally caused roots can be shut off. That is the external argument need not be projected in the syntax but its presence is felt because of our encyclopedic knowledge about the roots. Notice 10

that this is impossible in the verbal domain. Otherwise we would expect to find anticausative syntax with these roots, which we do not find. There are, however, two other domains in which we find something similar, the domain of adjectival passive formation and the domain of middle formation (Doron & Rappaport-Hovav 1991 and references therein).

•

(35)

a. b.

The carpet is destroyed This carpet destroys easily

Middle sentences and adjectival passives give us the feeling that an external argument is present, however this does not manifest itself syntactically. •

What are the similarities between the three constructions?

They all allow root semantics to be accessible via encyclopedic knowledge without the syntactic projection of an external argument. The three constructions are similar in that the event variable is not directly linked to Tense, but it is linked to a special kind of Aspect: genericity in the case of middles, stative Aspect in the case of adjectival passives (Embick 2004, Alexiadou & Anagnostopoulou 2007). 9 We thus need to search for the counterpart of this special kind of Aspect in the nominal domain. •

What is the trigger that allows encyclopedic knowledge to come into linguistic interpretation without having been represented syntactically in nominalization? We hypothesize that this relates to the mass noun interpretation these nominalizations have. Mass noun formation is part of the mechanisms languages can use when a generic interpretation is intended similarly to middle sentences, as is discussed in the semantics literature (see e.g. the contributions in Carlson & Pelletier 1995 and the discussion in the Papafragou's 1996 overview).

5. Further (open) questions • • •

In the nominal domain external arguments introduced by PPs do not make a distinction as to the preposition introducing them. Languages select for one of the two possible prepositions. German durch, English by, Greek apo. Why? How do nominalizations in other languages behave, which have a similar morphological distinction with anticausatives (e.g. Greek and the Romance languages)? What is the reason for the restriction on the external theta role that we find in some passives (Greek, Hebrew, German –en nominalization)?

9

In middles this violation of the semantic requirement on transitivity (that we find with agentive and externally caused roots) is rescued in the absence of thematic Voice by the presence of genericity. We are allowed to form anticausatives of externally caused roots if and only if genericity is present. Schäfer (2006) argues that the evaluation of genericity together with our world knowledge on externally caused roots brings about the agentive semantics. (i) a. *Der Präsident ermordet sich the president murders REFL b. Der Präsident ermordet sich nicht so leicht the president murders REFL not so easily

11

References Alexiadou, A. 2001. Functional structure in nominals: nominalization and ergativity. Amsterdam: John Benjamins. Alexiadou, A. & E. Anagnostopoulou. 2007. Structuring participles. To appear in Proceedings of WCCFL 26. Alexiadou, A., E. Anagnostopoulou & F. Schäfer 2006. The properties of anticausatives crosslinguistically. In: M. Frascarelli (ed.), Phases of Interpretation. Berlin: Mouton. Bhatt, R. & D. Embick. In progress. Causative Derivations in Hindi. Ms., University of Massachusetts at Amherst and University of Pennsylvania. Borer, H. 2001. The forming, the formation and the form of nominals. Handout USC. Carlson, G. & F. Pelletier. eds. 1995. The generic book. University of Chicago Press. Chomsky, N. 1970. Remarks on nominalization. In R. Jacobs, & P. Rosenbaum (eds.) Readings in English Transformational Grammar. Waltham (Mass.): Ginn & Co., 184221. Doron, E. & M. Rappaport-Hovav. 1991. 'Affectedness and Externalization'. Proceedings of NELS 21: 81-94. Eckardt, R. 2001. “Reanalysing selbst” Natural Language Semantics 9: 371 - 412. Ehrich, V. 1991. Nominalisierungen. In A. von Stechow & D. Wunderlich (eds.) Semantik; Ein Handbuch der internationalen Forschung. Berlin: Mouton de Gruyter, 441-458. Ehrich, V. & I. Rapp. 2000. Sortale Bedeutung und Argumentstruktur: ungNominalisierungen im Deutschen. Zeitschrift für Sprachwissenschaft 19: 245–303. Embick, D. 2004. On the structure of resultative participles in English. Linguistic Inquiry 35: 355- 392. Grimshaw, J. 2000. Argument Structure. Cambridge Mass.: MIT Press. Harley, H. & R. Noyer. 2000. Formal vs. Encyclopedic properties of vocabulary: evidence from nominalization. In B. Peters (ed.) The Lexicon-Encyclopedia Interface. Elsevier Press. Haspelmath, M. 1993. More on the typology of inchoative/causative verb alternations. In Causatives and Transitivity, Bernard Comrie, and Maria Polinsky (eds.), 87-120. Amsterdam: John Benjamins. Hole, D. 2006. Agentive selbst and the identity function. Ms., University of München. Kaufmann, I. 2003. Infinitivnominalisierungen von reflexiven Verben: Evidenz gegen Argumenstrukturvererbung? In C. Maienborn (Hrsg.) Asymmetrien: Festschrift für Ewald Lang. Tübingen: Stauffenburg, 197-226. Kratzer, A. 1994. The Event Argument and the Semantics of Voice. Ms., University of Massachusetts, Amherst. Levin, B., & M. Rappaport-Hovav. 1995. Unaccusativity. At the Syntax-Lexical Semantics Interface. Cambridge/Mass.: MIT Press. Marantz, A. 1984. On the Nature of Grammatical Relations. Cambridge/Mass: MIT Press. Marantz, A., 1997. No escape from syntax: Don’t try morphological analysis in the privacy of your own lexicon. In University of Pennsylvania Working Papers in Linguistics, Vol. 4.2, Alexis Dimitriadis, Laura Siegel, Clarissa Surek-Clark, and Alexander Williams (eds.), 201-225. Philadelphia. Marantz, A. 2005. Handout: Objects out of the lexicon: objects as event, presented at the University of Vienna. Papafragou, A. 1996. On generics. UCL Working papers in Linguistics 8. Ramchand, G. 2006. First phase syntax. Ms., University of Tromsoe. Schäfer, F. 2006. Middles as Voiced Anticausatives. In Proceedings of NELS 37, UIUC, Urbana-Champaign. Schäfer, F. In preparation. External arguments in change of state contexts: on the nature of anticausative morphology. Ph.D. diss, Universität Stuttgart. 12