24ne-raw [⌂ EvenYrbk]

Mark Newson

‘What do you get if you cross ergative with accusative?’ The Hungarian Causative*

1. Introduction Causative constructions typically have a verbal predicate and its arguments with an added element which introduces both a causative meaning and a causer argument. This predicate and its arguments can be called the base of the causative construction, encoding what would, in a non-causative, be a complete clause. Within the causative construction, however, this is realised as a phrasal complement of the CAUSE head and as such the construction is mono-clausal. Generally speaking, we can assume the following structure:1 (1)

CP C

TP T

vP DP(causer) v’ vCAUSE vP DP(agent) v’ v

VP DP(theme) V

1

This paper forms a small part of a much larger body of work that I have been undertaking in association with Krisztina Szécsényi over the past five or so years. Although the paper has been written independently, it still owes a great deal to Krisztina’s input to the project in general. Much as I would also like to share responsibility for any errors in this paper with her, I’m afraid that I have to claim these as my own. There is debate over the exact nature of the heads which introduce agents and causers. Kratzer (1996) proposed that the ‘voice’ head is responsible for introducing the agent and Pylkkänen (2008) proposes that a high applicative head introduces the causer. I do not intend to enter this debate in this paper and will simply assume that some head is responsible for these arguments and label them with neutral ‘v’ and ‘v CAUSE’ symbols.

The Hungarian Causative 75

Whatever the actual nature of the ‘v’ heads introducing the agent and causer arguments (see footnote 1), I will assume that they share the property of being phase heads and thus their complements will be spell-out domains. This will have repercussions for the analysis I will propose later on. Languages differ in the details of how this structure is realised. The causative head can appear independently or incorporated as exemplified by the two Chichewa causative constructions in (2) (Baker 1988: 148): (2)

a

b

Mtsikana ana-chit-its-a kuti mtsuko girl AGR-do-make-ASP that waterpot ‘The girl made the water pot fall.’ Mtsikana anau-gw-ets-a mtsuko girl AGR-fall-made-ASP waterpot ‘The girl made the waterpot fall.’

u-gw~e AGR-fall-ASP

Many languages have a causative where the causative head is realised only as a bound morpheme with the verb incorporated. Hungarian opts for this type:2 (3)

a

b

Az orvos meg-vár-at-t-a Mary-t the doctor PERF-wait-CAUS-PAST-3S Mary-ACC ‘The doctor made Mary wait.’ Az orvos meg-it-at-t-a Mari-val a főzet-et. the doctor PERF-drink-CAUS-PAST-3S Mary-INST the potion-ACC ‘The doctor made Mary drink the potion.’

As can be seen in (3), the base can either be intransitive or transitive. When it is transitive, the base subject behaves like an object and has accusative case. With a transitive base, however, the base object maintains its expected accusative case but the base subject takes instrumental case. Focussing on the case assignment within the base of the causative, this looks to have something in common with an ergative case system in which the transitive subject is assigned a case different from that assigned to the intransitive subject and the object, as exemplified in the following examples from Basque (Sanesteban et al. 2011: 1):3

2 3

The form of the causative morpheme depends on the base verb, partly due to vowel harmony and partly to lexical conditioning. The morphologically unmarked case assigned to the intransitive subject and object is traditionally called absolutive. However, in its tendency to be morphologically unmarked and the fact that it is generally applied to those DPs that do not have the distinct case (accusative in some languages, ergative in others), it has much in common with the

The Hungarian Causative 76

(4)

a

b

Pirata-Ø abiatzen da. pirate-NOM depart aux ‘The pirate departs.’ Medikua-k pirata-Ø beldurtzen du. doctor-ERG pirate.NOM frighten aux ‘The doctor frightens the pirate.’

This is not quite an accurate comparison, however, as in the causative base, both DPs have a case which is distinct from the subject. This might be compared to what is known as the tripartite system, as exemplified by Nez Perce (Baker 2015: 22, from Rude 1986: 126-7): (5)

a

b

Hi-páay-na háama. 3S-arrive-ASP man.NOM ‘The man arrived.’ Háama-nm hi-néec-‘wi-ye wewúkiye-ne. man-ERG 3S-PO-shoot-ASP elk-ACC ‘The man shot the elk (PL).’

In this system, none of the DPs share a case and both arguments of the transitive verb have distinct morphologically marked cases. In this brief paper, I will argue that the assignment of case we see in the Hungarian causative construction has things in common with both the accusative and ergative systems. However, despite appearances, it is not a tripartite system. In section 2, I will discuss Baker’s (2015) analysis of a very similar causative construction in Sakha, in which he proposes that ergative-like case assignment takes place within the causative base and an accusative system operates at the clause level. While this forms the basis for the analysis I will propose for the Hungarian causative, I will argue, in section 3, that the two languages are not identical in terms of how case is assigned within their causative constructions. Indeed, the Hungarian facts present a problem for Baker’s assumptions. Adopting proposals by Newson & Szécsényi (2022), originally designed to account for the interaction of case assignment and movement and which extend Baker’s theory, I show that these problems can be satisfactorily addressed in a way that is able to include both the Sakha and Hungarian case systems. Thus, the investigation both confirms and extends Baker’s approach to case assignment.

nominative case of languages with accusative objects. For this reason, many linguists refer to both as nominative.

The Hungarian Causative 77

2. Causatives in Sakha Hungarian isn’t the only language to have a differentiating case assigned to the base transitive subject of a causative. Baker (2015) discusses Sakha, a Turkic language spoken in Siberia, which displays virtually identical phenomena to Hungarian (Baker, 2015: 132): (6)

a

b

Sardaana Aisen-y yta(a)-t-ta Sardaana Aisen-ACC cry-CAUS-PAST.3SS ‘Sardaana made Aisen cry.’ Misha Masha-qa miin-i sie-t-te Misha Masha-DAT soup-ACC eat-CAUS-PAST.3SS ‘Misha made Masha eat the soup.’

The obvious difference between the Sakha and Hungarian causative constructions is that in Sakha the base subject of the transitive verb is assigned dative rather than instrumental case, as indicated by the fact that this is the form assigned to goals in the language (Baker, 2015: 132):4 (7)

Min Masha-qa kinige-ni I Masha-DAT book-ACC ‘I gave Masha the book.’

bier-di-m give-PAST-1SS

Baker analyses the case assignment within the Sakha causative as ergative-like, in that the base subject of the transitive is assigned a differentiating case. Based on Marantz’s (1991) Dependent Case Theory (DCT),5 Baker assumes the following parameterised principle of dependent case assignment: 4

5

A reviewer asks if dative is structural in Sakha in other constructions. Baker claims that it is structural inside VP arguing on the basis that Sakha has no double accusative structures. Thus, double object constructions always render the goal as dative. However, these observations do not exclude the possibility that the dative on the goal is not an inherent case. Baker and Vinokurova (2010) argue that the dative associated with the goal is structural on the basis of the observation that a goal will be accusative when accompanied by a CP, as with the verb ‘promise’ for example. However, if an argument raises out of the CP into the same domain as the goal, the goal becomes dative. This is fairly strong evidence that dative is a structural, not an inherent case in Sakha. Marantz proposes that case is assigned to eligible DPs in a construction in the following order: a) lexical case b) dependent case c) unmarked case d) default case

The Hungarian Causative 78

(8)

If XP bears c-command relationship Y to ZP in local domain WP, then assign case V to XP.

In order for this principle to operate, it is essential that there be a c-command relationship (Y) between two potential case recipients, i.e. dependent case can only be assigned in a transitive context, the core assumption of DCT. Standardly, any DP which is not assigned dependent case is assigned unmarked case. Thus, the solitary DP in the intransitive context and the DP which is not assigned dependent case in the transitive context will both be assigned the unmarked case. The value for Y is parameterised, allowing for dependent case to be assigned to the ‘c-commanding’ (structurally higher) or ‘c-commanded’ (structurally lower) phrase. In this way, the distinction between the accusative and ergative systems is achieved as either the higher or lower phrase will be assigned the same (unmarked) case as that in an intransitive context while the other gets the differentiating case. A second parameter in (8) concerns the identity of the local domain WP. Baker claims that the domains relevant for limiting the operation of dependent case assignment are the spell-out domains of Phase Theory (Chomsky 2000, 2001). Thus, TP is the relevant domain for the assignment of dependent case at the clause level, with the subject and the object being the two potential recipients of the dependent case. NP within the DP is also a case domain, though we will not touch on this in this paper (see Newson & Szécsényi 2020 for some discussion of case assignment within the Hungarian DP from this perspective). Crucially for Baker’s account of the Sakha causative, the VP is also a case domain. The fact that (8) is given with respect to a domain means that in principle a language can select a different dependent case assignment (to the ccommanding or the c-commanded DP) in different domains. This is the crux of Baker’s analysis of the Sakha causative which claims that while Sakha adopts an accusative system in the TP, it has an ‘ergative’ system in the VP.6 Within the causative, the VP contains the arguments of the base verb. If this is

6

Once a DP is assigned a case, it becomes inactive in the system. Lexical case is assigned to specific arguments of specific predicates and dependent case can only be assigned if three are two active DPs in a construction. Which of these DPs dependent case is assigned to is a language specific choice. All DPs which have not been assigned a lexical or a dependent case will be assigned unmarked case. Essentially, the ergative pattern emerges when dependent case is assigned to the higher of the two arguments and the accusative pattern when it is assigned to the lower one.

The Hungarian Causative 79

transitive, the c-commanding argument (the base subject) is selected to receive the dependent case.7 The third parameter, which is relevant for the distinction between Sakha and Hungarian, concerns ‘case V’. The parameter allows for a different case to be selected as the assigned dependent case for a particular case domain. So, while Sakha selects accusative as its dependent case at clause level, it selects dative as the dependent case assigned within the VP. The obvious difference between Sakha and Hungarian is that Hungarian selects instrumental as its dependent case within the VP. The fact that the Sakha object is assigned accusative case obviously means that it adopts the accusative pattern of dependent case assignment at clause level. However, if the object is in the VP, which is a case domain, it cannot be considered to be in a ‘transitive’ context with the subject, which would sanction dependent case assignment. Baker argues that this is indeed the case, as the object is only assigned accusative if it raises out of the VP into the same domain as the subject. This happens only if the object is definite; the indefinite object receives unmarked case: (9)

a

b

Masha salamaat-y türgennik sie-te. Masha porridge-ACC quickly eat-PAST.3SS ‘Masha ate the porridge quickly.’ Masha türgennik salamaat sie-te. Masha quickly porridge eat-PAST.3SS ‘Masha ate porridge quickly.’

As Baker points out, the order of the accusative phrase with respect to the VP adverb, suggests that it is outside the VP in (9a) but inside it in (9b), the movement having to do with the definite nature of the accusative marked one. When the object is within the VP, it is eligible only for unmarked case as it is the only DP within this domain. However, when it moves out of the VP, it is situated within the same domain as the subject and as Sakha is an accusative language in this domain, dependent case is assigned to the lower of the two DPs. Although Baker does not provide the data to show it, we can assume that the object in the causative shown in (6b) is outside the VP as it is definite. I assume that if it were indefinite, it would have unmarked case and remain inside the VP. Note that the dative DP precedes the definite object and so it too must be outside the VP (see (7)), in line with Baker’s claims as it is also definite. Baker does not discuss the issue of why the dative case is maintained after the

7

We will discuss the accusative assigned to the object shortly.

The Hungarian Causative 80

movement. However, I will discuss a possible approach to this issue when discussing a problem which arises with the Hungarian causative.8

3. The Hungarian causative As we have seen, the Hungarian causative construction has a lot in common with that of Sakha, the most obvious difference being the choice of the dependent case within the VP domain: dative for Sakha and instrumental for Hungarian. However, there is another not so obvious difference which has a far greater impact on the analysis of the Hungarian system and indeed has repercussions for the theory of case domains. Hungarian, again like Sakha, adopts an accusative case system at the clausal level, meaning that it selects to assign dependent case to the lower argument (the object). Therefore, the subjects of intransitive and transitive verbs get unmarked case: (10) a

b

Mari- sír-t- Mary-NOM cry-PAST-3S ‘Mary cried.’ meg-it-t-a a főzet-et Mari- Mary-NOM perf-drink-PAST-3S the potion-ACC ‘Mary drank the potion.’

Unlike Sakha, however, there is no reason to believe that the object has to move out of the VP to be in the same domain as the subject. While there are plenty of possible movements from the VP, Hungarian being a language in which word order is very much determined by such discourse notions as topic and focus, none of these affect the case born by any DP:9 8

9

There is another possible, but less attractive, solution that Baker could adopt which assumes that neither DP in the causative moves, though this would raise the issue of how they can avoid this, given their definiteness. If both DPs remain inside the VP, then they both must be assigned dependent case within this domain. This suggests that Sakha does not adopt the ergative strategy for case assignment, which would assign dependent case to the base subject and unmarked case to the object. Instead, the tripartite strategy which arises by allowing both DPs to be assigned dependent cases, must be assumed for the Sakha VP. I consider this unappealing, however, as the tripartite strategy is very rarely adopted and one would need rather more direct evidence to warrant assuming it in any particular case. Much rests on the empirical question of what happens with indefinite DPs in the causative. Unfortunately, I do not have access to that data. Hungarian does not show any differential object marking effects in terms of the case assigned to definite and indefinite DPs, as happens in Sakha. It does however show a

The Hungarian Causative 81

(11) a

b

a főzet-et meg-it-t-a Mari- the potion-ACC Mary-NOM perf-drink-PAST-3S ‘The potion, Mary drank.’ a főzet-et it-t-a meg Mari- Mary-NOM the potion-ACC drink-PAST-3S perf ‘It was the potion that Mary drank.’

It seems, therefore, that the object’s case is determined before it undergoes any movement, i.e. within the VP. In order to account for languages in which objects within the VP are still able to interact with subjects to determine the assignment of dependent case, Baker proposes a separate parameter which affects the status of the VP as a case domain. His claim is that languages either have ‘hard’ or ‘soft’ VP case domains. Within a hard domain, DPs assigned a case become inactive and play no further role in determining case assignment in other domains. This means that they cannot be assigned a case determined by the conditions holding in a higher domain and neither can they play a role in determining those conditions, governing the case assignment to the DPs in that domain. Although soft domains still count as spell-out domains, the DPs within them remain active into the following phase. Therefore, they are able to be assigned a case in that phase and can interact with other DPs in the higher domain to allow dependent case to be assigned. It appears that the Hungarian VP is soft. This conclusion is problematic in the case of the causative structure. If VP internal DPs remain active into the clause phase, the assignment of instrumental case to the base subject could be affected. It appears, then, that the VP acts as a hard domain for the instrumental case assignment but as a soft domain for the accusative. This is not something that (8) can be parameterised to account for. Note that we cannot overcome this problem by assuming that the Hungarian VP has a tripartite system with dependent case being assigned to both base arguments in the causative. If the Hungarian VP were hard, we would expect unmarked case to be assigned to the object in transitive constructions and causatives with an intransitive base. The fact that these DPs are assigned accusative case means that the VP must be soft, and therefore the problem persists. definiteness distinction in terms of the agreement paradigm on the verb. When the verb has a definite object, the paradigm of subject agreement used differs from that when the object is indefinite. While this could be taken to indicate that the two objects occupy different structural positions, there is no independent empirical evidence to support such a hypothesis. The fact that the indefinite paradigm is the same as the one used with intransitive verbs supports the claim that indefinite objects are invisible for agreement purposes (see Newson & Szécsényi 2020).

The Hungarian Causative 82

To overcome the problem, I turn to a proposal made by Newson & Szécsényi (2022) to account for why some movements ‘change’ the case of a DP while others maintain it. This developed the notion of a soft domain by introducing the idea that the assignment of unmarked case is delayed in soft domains and is only actualised in a hard domain. Assuming that the assignment of case is what deactivates a DP, this accounts for why certain DPs in a soft VP remain active while others do not. Dependent case is assigned in a soft domain, deactivating that DP. Those earmarked for unmarked case will face delayed case assignment and thus remain active. Let us take some cases and provide a more detailed analysis to show how this works. Let us start with a simple transitive verb. The first merger involves the verb with its theme argument to form the VP and this becomes a case domain once the phase head, introducing the agent, is merged with it. At this point the theme is the only DP and hence unmarked case can be assigned to it. However, as the VP is a soft domain, indicated by underlining in (12), the assignment of this case is delayed: (12)

[vP v [VP DPTHEME V]]

Theme earmarked for unmarked case

The derivation continues with the merger of the agent and the tense. But it is only after the complementiser is merged that the next case domain is triggered. This will be a hard domain, indicated by bold underlining in (13), so all cases will need to be assigned. Both the theme and the agent are active at this point and hence dependent case can be assigned. In an accusative system, this will be assigned to the object, overriding the unmarked case it would have been assigned had the VP not been soft. Unmarked case will be assigned to the subject: (13) [CP C [TP T [vP DPAGENT v [VP DPTHEME V]]]] Theme assigned accusative; Agent assigned unmarked Assuming that the causative head is a phase head and moreover that it, like the head which introduces the agent, produces a soft case domain, exactly the same will hold in the causative with an intransitive base. Unmarked case assignment to the base subject will be delayed until the hard domain of the clause has been constructed, at which point both the causer and the base subject are active and accusative case will be assigned to the lower of these.

The Hungarian Causative 83

(14) [vP vCAUSE [vP DPAGENT V [VP …]]] [CP C [TP T [vP DPCAUSER vCAUSE [vP DPAGENT v [VP …]]]]]

Agent earmarked for unmarked Agent assigned accusative; Causer assigned unmarked

In the causative with a transitive base, things are a little different. It starts similarly to the derivations detailed above up to the merger of the agent. At this point, the object and the agent are still active as the object is in a soft domain and the agent is not in any domain as yet: (15) [vP DPAGENT v [VP DPTHEME V]] Theme earmarked for unmarked case; Agent undetermined The CAUSE head is then merged introducing a soft domain, this time including the agent, as this has a similar nature to the head introducing the agent. As this domain now contains two active DPs, dependent case can be assigned. In Hungarian, in the context of this particular domain created by the introduction of the CAUSE head, the higher of the two is selected and this is realised as instrumental: (16) [vP vCAUSE [vP DPAGENT V [VP DPTHEME V]]] Agent assigned instrumental; Theme earmarked for unmarked The agent is therefore deactivated. The theme is once again earmarked for unmarked case, which will be delayed until a hard domain has been formed. Only once the complementiser is merged is this domain created and at this point the causer and the base object are both active and available for case assignment. (17) [CP C [TP T [vP DPCAUSER vCAUSE [vP DPAGENT v Agent deactivated; Theme assigned [VP DPTHEME V]]]] accusative; Causer assigned unmarked As in all Hungarian transitive clauses, the result is that the lower argument is assigned the dependent and the higher one the unmarked case. This analysis does not affect Baker’s proposals for Sakha. As the Sakha VP is a hard case domain, unmarked case assignment will not be delayed and therefore will be assigned to the object unless it moves out of the VP. The base subject, however, will be assigned dependent case whether it moves out of the

The Hungarian Causative 84

VP or remains within it. Being deactivated by this case assignment, it plays no further role in determining the case assigned to other DPs in any domain. The overall effect of this analysis is that the Hungarian causative VP acts as a domain for one of the DPs it contains, the one assigned dependent instrumental case, but not for the other, since the assignment of unmarked case in a soft domain is delayed until a hard domain is formed. The result is that the arguments within the base verbal structure are both assigned a dependent case, but not in the way it is done in a tripartite system. Essentially, what we have is that an accusative case system is superimposed on top of a structure in which an ‘ergative’ case system has already been applied.

4. Conclusion The case assignment patterns we see in the Hungarian causative construction very much support the general theory proposed by Baker, demonstrating parametric variations in exactly the areas this theory predicts. It provides evidence that the choice of dependent case realisation is not only language particular but also domain particular, within a specific language. It also shows that the choice of case system, ‘high dependent case’ vs. ‘low dependent case’, varies not just from language to language but across different domains within the same language. At the same time, this paper has shown that this construction presents challenges for the theory which necessitate development in certain parts. In many ways, the distinction between soft and hard case domains is one of the less attractive aspects of Baker’s work as the exact nature of this distinction is poorly understood and remains fairly stipulative. It can be hoped that the kind of development proposed in this paper to the notion will enable a fuller understanding and eventually a better integration into the theory of case assignment. It seems to me that we are only at the start of this road, but the proposals made here perhaps represent one step along it. References Baker, Mark. C. 1988. Incorporation: A theory of grammatical function changing. Chicago: University of Chicago Press. Baker, Mark. 2015. Case: its principles and parameters. Cambridge: Cambridge University Press. https://doi.org/10.1017/CBO9781107295186 Baker, Mark C. and Nadya Vinokurova. 2010. Two modalities of case assignment: case in Sakha. Natural Language and Linguistic Theory 28. 593–642. https://doi.org/10.1007/s11049-010-9105-1 Chomsky, Noam. 2000. Minimalist inquiries: The framework. In: Roger Martin, David Michaels and Juan Uriagereka (eds.). Step by Step. Cambridge, Mass.: MIT Press. 89–155.

The Hungarian Causative 85 Chomsky, Noam. 2001. Derivation by phase. In: Michael Kenstowicz (ed.). Ken Hale: A life in language. Cambridge, Mass.: MIT Press. 1–52. https://doi.org/10.7551/mitpress/4056.003.0004 Kratzer, Angelika. 1996. Severing the external argument from the verb. In Johann Rooryck and Laurie Zaring (eds.). Phrase Structure and the Lexicon. Dordrecht: Springer. 109–137. https://doi.org/10.1007/978-94-015-8617-7_5 Marantz, Alec. 1991. Case and Licensing. In ESCOL 91: Proceedings of the eighth Eastern States Conference on Linguistics. Columbus: Ohio State University. 234-253. [Reprinted in Eric Reuland (ed.). Arguments and Case: Explaining Burzio’s generalization. Amsterdam: John Benjamins Publishing Co. 11 – 30.] Newson, Mark and Krisztina Szécsényi. 2020. Dative as the unmarked unmarked case in Hungarian. In Markéta Ziková and Marcin Wągiel (eds.). Linguistica Brunensia, 68.2. Brno: Masarykova univerzita. 27– 44. https://doi.org/10.5817/LB2020-2-3 Newson, Mark and Krisztina Szécsényi. 2022. Domains of case changing and case maintaining movements. Linguistics Beyond and Within 8. Lublin: Faculty of Humanities, John Paul II Catholic University. 152–166. https://doi.org/10.31743/lingbaw.14963 Pylkkanen, Liina. 2008. Introducing arguments. Cambridge: MIT Press. https://doi.org/10.7551/mitpress/9780262162548.001.0001 Rude, Noel. 1986. Topicality, transitivity, and the direct object in Nez Perce. International Journal of American Linguistics 52. 124–153. https://doi.org/10.1086/466009 Santesteban, M., Pickering, M., and Branigan, H. 2011. Exploring thematic role and case marking effects in structural priming. Poster presented at the annual CUNY conference on Human Sentence Processing, Stanford, USA. Mark Newson Eötvös Loránd University ORCHID 0000-0002-7127-6448 newson.mark@btk.elte.hu