22ns-raw [⌂ EvenYrbk]

Mark Newson and Krisztina Szécsényi

Not so Exceptional Case Marking in English*

1. Exceptional Case English Exceptional Case Marking (ECM) constructions involve an accusative subject in a non-finite clause: (1)

a

John expects [me to take the blame]

Generative tradition has it that what is ‘exceptional’ here is that the infinitival subject bears a case which is licensed in the way that objects have their cases licensed, i.e. by the relevant preceding verbal element1. Thus, these subjects display certain objective properties. Nominative subjects have their case licensed from a functional element within the same clause, which, being more frequent cross-linguistically, is assumed to be non-exceptional. In the present paper we will present another analysis, from within Dependent Case theory (DCT), in which the ‘accusative’ case assigned to the subject of the infinitival clause is not treated as different to the case assigned to any subject and is therefore not at all exceptional. It will be claimed that all subjects in English are assigned unmarked case, but for a small number of elements (five pronouns) this gets realised differently depending on context. But even from this perspective, an ‘accusative’ realisation is not exceptional. Indeed, it could be argued that it is the ‘nominative’ realisation that is the exception, as in most contexts the realisation of a pronoun with unmarked case is with the ‘accusative’. In the preceding paragraph the terms ‘accusative’ and ‘nominative’ were placed within inverted commas to indicate that we will be taking a special view

1

We would like to thank Polina Pleshak for helpful comments on a draft of this paper, Irina Burukina for discussion on its main ideas and the audiences of various talks at which some bits of the paper have been presented for their questions and comments. We take responsibility for any remaining errors. We are being somewhat equivocal in the use of terms here to avoid going into unnecessary details. As we will not be basing our analysis on Standard Case Theory, we do not want to take a stance concerning which head is involved with accusative case (V, v, Agro, etc.) and we adopt the neutral term ‘licensing’ to refer to the relationship between the head and the DP bearing the case to cover specific ideas such as ‘case assignment under government’, ‘feature checking’, ‘feature valuation’, etc.

Not so exceptional case marking in English 51

of such terms. Using such names has often been assumed to be a positive way to compare and contrast different case systems in the world’s languages. With their use, for example, we can see similarities between the Hungarian subject of a finite clause, which carries no case morphology, and the equivalent element in Japanese, which is marked for case. This is because both languages have a nominative-accusative system, assigning one case (morphologically unmarked in Hungarian but marked in Japanese) to grammatical subjects and a different one to the object (marked in both languages)2. (2)

a János- meghív-t-a János-NOM invite-PAST-3S. ‘John invited Mary.’ b Mari- alsz-ik. Mary-NOM sleep-3S. ‘Mary is sleeping.’

Mari-t. Mari-ACC

(3)

a Inu-ga mizu-o no-mu. dog-NOM water-ACC drink-PRES ‘The dog drinks water.’ b Shojo-ga aru-ku girl-NOM walk-PRES ‘The girls walks.’

In turn, we can then contrast these languages with those which have an ergativeabsolutive system, such as Basque, where the subject of the intransitive clause is assigned the same case as the object (absolutive), while the subject of the transitive gets another case (ergative). (4)

2

a Jon-ek sagar bat jan du. John-ERG apple a.ABS eat have.(3S.ABS.)3S.ERG ‘John has eaten an apple.’ b Mahai-a apurtu da. table-DET.ABS break be.3S.ABS ‘The table broke.’

Baker (2015) claims that Japanese is a marked nominative language, which in his system is a distinct case system characterised by a negative condition on the assignment of dependent case (assign dependent case to a DP which is NOT c-commanded by another DP in the same domain). In such a system, ‘nominative’ is the dependent case and ‘accusative’ is unmarked. However, while there are languages in which nominative is morphologically marked and accusative is unmarked, in Japanese both nominative and accusative are marked.

Not so exceptional case marking in English 52

However, we will argue that the use of such case names can also be misleading in certain instances. English offers a good example. Calling forms such as he, she, I, etc. nominative and those such as him, her and me accusative predisposes us to assume that English is similar to Hungarian in having a nominative-accusative system. However, we will argue that this is not the case and that English is more like Chinese in terms of its case system, having a neutral case system in which subjects and objects are assigned the same unmarked case. (5)

a Ni/Zhangsan kanjian wo/Lisi You/Zhangsan see me/Lisi ‘You/Zhangsan see(s) me/Lisi.’ b Wo kanjian Zhangsan I see Zhangsan. ‘I see Zhangsan.’ c Wo/Lisi xing le I/Lisi woke.up asp. ‘I/Lisi woke up.’

Our claim is that English and Chinese differ only in terms of how unmarked case is realised in different contexts: Chinese realises unmarked case in all contexts uniformly whereas English, for reasons to do with its historical development, realises unmarked case on certain pronouns differently in different contexts. In the next section we will introduce DCT and discuss the difference between case assignment and case realisation that it allows. Following this we will present evidence that the subject of the infinitive is assigned unmarked case rather than the dependent case expected under more standard assumptions. We will extend the analysis to cover all objects too, meaning that English is a language which assigns no dependent case to either subjects or objects and therefore has a neutral case system, as defined within DCT.

2. Dependent Case Theory and the difference between case assignment and case realisation Dependent Case Theory (Marantz, 1991; Baker, 2015) is concerned with the distribution of morphological case rather than the licensing of argument DPs. It further differs from Standard Case theory in that it assumes that the conditions of case assignment are determined by the relationships holding between DPs rather than between a DP and a head. It classifies cases into four types: lexical;

Not so exceptional case marking in English 53

dependent, unmarked and default. We will be concerned mainly with dependent and unmarked case in this paper. A core component of DCT sets out the conditions under which dependent case is assigned. In general terms this can be stated as follows: (6)

If two DPs, α and β, are contained in the same domain and α ccommands β, then dependent case can be assigned to: a. α b. β (choice of a. or b. parameterised)

The key points here are that dependent case assignment is dependent on the presence of at least two c-command related DPs within a limited section of a structure, the domain, which we will discuss a little later. This can be assigned to the higher (c-commanding) or the lower (c-commanded) DP, depending on the parametric choice of the language, perhaps relative to the domain. So, it would be possible for a language to assign a high dependent case in one domain and a low dependent case in another. Unmarked case is assigned to a DP which is not assigned a dependent case3. This means that in intransitive contexts the lone DP will be assigned unmarked case, alongside the DP not assigned dependent case in transitive contexts: the subject in accusative languages and the object in ergative languages. Baker (2015) adds two extra parameter values which allow dependent case to be assigned to both DPs present in a domain, giving rise to a tripartite case system, which will not be concerned with here, or to neither of them. This latter option produces the neutral system in which there is no case distinction between subjects and objects within the clause. Turning to the domain, we can see from (6) that one of its functions is to limit the DPs which determine the assignment of dependent case. So, in (7), where γ is a domain, dependent case could be assigned in a) but not b): (7)

a b

[ … [γ DP … DP ]…] [ DP … [γ DP …] …]

However, domains are also the determining factor for the assignment of specific cases. This is because different dependent and unmarked cases are defined relative to specific domains. For example, at the clause level nominative might

3

Technically, unmarked case is assigned to DPs which are not assigned either lexical or dependent case, but we ignore the issue of lexical case here.

Not so exceptional case marking in English 54

be the unmarked case, while in the nominal domain, the unmarked case could be genitive, as suggested by Marantz (1991). Baker posits that the domain relevant for case assignment is the same thing as the Spell-Out domain relevant for phase theory (Chomsky 2000, 2001). Thus, IP, VP and NP4 are taken as case domains. Case assignment takes place at SpellOut and possibly interacts with movement in that the case a DP bears can be determined after it has moved rather than in its base generated position. The view that VP is a domain is problematic for some languages in which subjects and objects interact to allow the assignment of dependent case. If the object is inside VP and the subject external to it, they should be in different domains unless the object raises. Indeed, Baker uses the fact that object raising does affect case assignment in some languages, Sakha for example, to support his view that VPs are case domains. For those languages where there is no reason to assume object raising but in which subjects and objects still interact, Baker suggests that VPs can be ‘soft’ domains, meaning that certain DPs remain active in these domains even after Spell-Out. In the present paper, we will make crucial use of the assumption that VP can be a hard case domain. 2.1. The status of cases It will be useful at this point to consider the status of the notions ‘dependent case’ and ‘unmarked case’ within the theory and how they relate to specific cases such as nominative, accusative, dative, etc. Anticipating conclusions reached a little later, we will refer to the former as ‘cases’ and the latter as ‘case forms’. Both Marantz and Baker state that what is actually assigned are case forms. For Marantz, these amount to features assigned to case morphemes by certain heads or combinations of heads. Baker sets out the general condition on the assignment of dependent case in the following way (p. 79): (8)

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

As this concerns the assignment of dependent case, Baker’s reference to ‘case V’ must mean different case forms. It seems, therefore, that the notions ‘dependent’ and ‘unmarked’ case are to be seen as labels for types of different case forms. 4

More precisely, the nominal case domain will be whatever phrase is in the complement position of the determiner, under the assumption that determiners are phase heads. This could be any of a number of things, depending on what functional structure and their organisation one assumes. The issue is not important for the present paper.

Not so exceptional case marking in English 55

However, there is another possible view which seems to exist in the theory. Marantz states (p. 25): “ACC is the name for the dependent case that is assigned downward … ERG is the dependent case assigned upward …” In other words, accusative and ergative are what we call dependent case assigned to different positions. This is opposite point of view to the one presented above. Here, it is dependent case which is the main concern of case assignment and case forms are merely labels for the different instantiations of this, determined by the conditions of its assignment5. If we adopt this perspective, there are only two structural cases assigned in the syntax. The different case forms would then have to be seen as different realisations of these. This would work under a late insertion approach whereby the form selected to realise dependent and unmarked cases is determined by the context in which the DP is situated (Halle and Marantz 1993). In effect, this approach separates the two functions attributed to domains: that of determining the conditions under which case is assigned and that of determining what specific form DPs bearing those cases take. The former is a purely syntactic issue and therefore it comes as no surprise that it is constrained by syntactically relevant notions such as Spell-Out domains. The latter is a post syntactic matter restricted by some notion of context, which may or may not be structurally defined. Importantly, the latter does not have to correspond to the Spell-Out domain. One argument in favour of this approach is that the domains of case assignment are not always identical to the contexts of case form realisation. We will give an example of this a little later, after considering another argument which points to the same conclusion. Baker considers the issue of marked nominative and marked absolutive systems, which show the same case alignments as accusative and ergative systems but differ in terms of which case is morphologically marked. For example, in a marked nominative system, the subjects of both transitive and intransitive constructions bear the same morphological case and the object is typically unmarked. To account for this pattern, he introduces extra conditions for the assignment of dependent case (p. 93): 5

A position similar to this is already assumed by linguists who reject the distinction between ‘nominative’ and ‘absolutive’, claiming that these constitute a single case, usually termed ‘nominative’, assigned to the typically morphologically unmarked DP which is not assigned either accusative nor ergative (e.g. Bitner and Hale 1996). From this perspective ‘absolutive’ is merely the name given to nominative case in the ergative system.

Not so exceptional case marking in English 56

(9)

a Assign NP1 marked nominative if there is no other NP, NP2, in the same TP as NP1 such that NP2 c-commands NP1. b Assign NP1 marked absolutive if there is no other NP, NP2, in the same TP as NP1 such that NP2 is c-commanded NP1.

This not only increases the number of conditions under which case assignment is determined, but it also adds a different kind of determining principle, as the principles in (9) involve a negative c-command condition as opposed to the simple c-command conditions on the assignment of accusative and ergative. Clearly, by these definitions, marked nominative/absolutive cases in these systems are dependent and accusative. Correspondingly, ergative are unmarked, not only morphologically but in terms case type as defined by the DCT system. They therefore radically differ from the accusative and ergative cases in the standard systems. Baker argues in favour of this approach, as opposed to one where there is a simple mismatch between syntactic and morphological unmarkedness, on the grounds that it allows the two to remain aligned. However, this argument is seriously undermined by the fact that elsewhere syntactic and morphological marking are not always aligned. For example, a language which Baker analyses is Tamil, an accusative language which follows the usual pattern in having a morphologically unmarked nominative subject. Yet he argues that genitive is the unmarked case in the nominal domain, despite the fact that this is morphologically marked. Newson and Szécsényi (2020) analyse Hungarian as having an unmarked dative subject in inflected infinitives, which differs from the unmarked nominative subjects of finite clauses by displaying overt dative morphology. Moreover, some of the examples of marked nominative languages that Baker himself discusses, specifically Japanese and Korean, also have marked accusative objects. If these languages assign dependent case to their subjects, then the object must be syntactically unmarked, despite being morphologically marked. It seems therefore that there is no advantage to be gained by complicating the grammar with additional case assignment principles, such as those in (9). Assuming that ‘marked nominative/absolutive’ cases are syntactically unmarked but realised through overt morphology does away for the need of negative c-command conditions on dependent case assignment, thus producing a simpler system. This state of affairs is also what might be expected in a system in which case assignment and case realisation are separate, as we are proposing. It is arguable that the connection between syntactic and morphological unmarkedness, although it may be a tendency crosslinguistically, is mainly a terminological one in DCT. If Marantz had chosen to call the case assigned to DPs which are not assigned dependent case something other than ‘unmarked

Not so exceptional case marking in English 57

case’, the expectation that it should be unmarked morphologically would have been diluted. 2.2. Assignment and realisation As an example of the system we have in mind, we will briefly review the analysis of Hungarian nominative-dative alternation in Newson and Szécsényi (2020). The data concern nominative and dative DPs in clausal and nominal environments. At the clause level, we find nominative subjects in finite clauses and dative subjects in inflected infinitive clauses: (10) a … hogy Péter- that

Peter-NOM

[ lát-ja

a

kutyá-t ] / [ táncol- ]

see-3SG the dog-ACC

dance-3SG

‘… that Peter sees the dog/dances’ b Nem szabad [ Péter-nek lát-ni-a not allowed

a kutyá-t

/ táncol-ni-a]

Peter-DAT see-INF-3SG the dog-ACC

dance-INF-3SG

‘Peter is not allowed to see the dog/dance’ Within the DP, the same alternation can be found on the possessor, depending on its position. Possessors in the lower position are nominative and in the higher position dative: (11) a

Péter-

kalap-ja / az

én

kalap-om

I.NOM

hat-1SG

Peter.NOM hat-3SG

the

‘Peter’s hat’

‘my hat’

b Péter-nek

a

kalap-ja

/ nekem a

Peter-DAT the hat-3SG

I.DAT

‘Peter’s hat’

‘my hat’

kalap-om

the hat-1SG

As demonstrated in (10), nominative and dative subjects follow the same unmarked pattern in appearing in both transitive and intransitive contexts. Clearly, one could not be analysed as dependent and the other unmarked. The fact that there is no semantic difference between these DPs indicates that one cannot be taken to be unmarked and the other lexical. The same conclusions carry over to the nominative and dative possessors. Assuming that nominative and dative cases are assigned under different conditions, Newson and Szécsényi argued that the domains of nominative case

Not so exceptional case marking in English 58

assignment must differ from those of dative case assignment. For nominative case assignment, the domain is the complements of C and D, identified as canonical heads of extended verbal and nominal projections. The unmarked dative appears not only in the contexts in (10) and (11), which are differentiated from the unmarked nominative domains in that they are the complements of the non-finite agreement, but also in a number of places, such as on certain nominal and adjectival predicates, which leads to the conclusion that the unmarked dative domain is to be seen as the elsewhere domain6. One consequence of this analysis is a proliferation of phases, some of which seem only relevant for case assignment. However, under the assumption that there is just one case assigned in these constructions, unmarked case, and that this is realised by different case forms, there is possibly no need to assume so many domains. We might assume that there is a single clausal and nominal phase, in line with standard assumptions, in which unmarked case is assigned. However, the contexts for realising this case are defined along different lines and hence different case forms appear. It can be claimed that unmarked case will be realised as nominative in a context associated with a ‘canonical head’ (HC), such as C and D, and dative elsewhere7. We propose the following realisation rules for unmarked case in Hungarian: (12) [unmarked]   / [HC —] [unmarked]  nAk elsewhere

3. ‘Accusative’ subjects in English Newson (2019) argued that the accusative subject of English acc-ing gerunds is an instance of unmarked case. Following Baker’s (2015) suggestion that the NP inside the DP is relevant for determining the possibility of assigning dependent case within the DP and Abney’s (1987) analysis of DP structure, in which the acc-ing gerund lacks an NP, the most coherent analysis is to take genitive as dependent and accusative as unmarked. 6

7

Polina Pleshak (p.c.) points out that the notion of an elsewhere domain suggests a link to default case. However, we take the view of Marantz (1991) that default case is different to other cases, appearing separately on the case hierarchy. In fact, how default case fits in with DCT assumptions is not well investigate. But see Newson (2018) where it is argued that default case occurs only in the absence of an assigned case and unmarked case is assigned. The notion of an elsewhere domain is strictly to do with assigned cases and is defined as the set of domains which remains after subtracting those domains identified for the assignment of specific cases (such as nominative). This account obviously needs further elaboration, but the present paper is not the place for this. Its purpose here is to exemplify how we propose to actualise the distinction between case assignment and realisation.

Not so exceptional case marking in English 59

This, then, raises the possibility that other instances of accusative subjects are unmarked rather than dependent. With this in mind, consider the accusative subject of English Small Clauses: (13) They considered [us aggressive] The standard view is that these accusative subjects should be analysed as within an extension of the domain of the governing verb, making the subject the grammatical object of the verb, as with ECM constructions. However, this is questionable on the same grounds that such an analysis cannot work for acc-ing gerunds, as can be seen by comparing (14) and (15). (14) a This policy made [them nostalgic for a more authoritarian leadership]. b [Them nostalgic for a more authoritarian leadership] is a situation to be avoided8. (15) a We applauded [them voting against the proposal]. b [Them voting against the proposal] was a setback for the government. In both cases, as these constructions can appear in subject positions themselves, it would be hard to account for their accusative subjects as extended objects of the lower verb from within any framework9. Specifically within DCT, the lack of a c-commanding DP in the b) examples makes the assignment of a low dependent case (i.e. accusative) hard to fathom. Equally as problematic are independent Small Clauses (elsewhere known as ‘Mad mag sentences’, Akmajian 1984)10:

8 9

10

The use of Small Clauses as subjects was pointed out in Stowell 1981. Polina Pleshak (p.c.) suggests that in some of these cases the accusative could be licensed from a c-commanding DP before movement takes place. However, in the examples given in (14) and (15) it is doubtful that the accusative subject originates in a position where it is ccommanded by a DP and there are other cases where the lack of any other DP demonstrates that this cannot be the source of a dependent accusative: i) Them running away was disrespectful. ii) Them free is worrying. Schütze (2001), assuming a left dislocation analysis for these constructions, argues that the accusative case on these DP is default. Newson (2018) offers counter arguments.

Not so exceptional case marking in English 60

(16)

a What, me worry? b Them give me a rise?! Never! c Him wear a tuxedo?! That, I will have to see!

The lack of a c-commanding DP for the accusative subject in all these constructions most naturally leads to the conclusion that they are instances of unmarked case assignment. 3.1. For infinitives As a step towards an analysis of ECM constructions, let us first turn our attention towards for infinitives, which share with them an accusative subject and an infinitive tense. One main difference between for infinitives and ECM constructions is that like gerunds and Small Clauses, the former can appear in both subject and complement positions: (17) a The government were anxious [for the bill to be passed quickly]. b [For the bill to be passed quickly] is important for the government. The role of the complementiser as the licensor for the accusative subject has traditionally been the assumed account of this observation. However, from a DCT perspective, the traditional analysis cannot be maintained. Having abandoned the use of head licensing, we cannot assume that the complementiser has any direct responsibility for the assignment of accusative in this construction. Given that the subject of the for clause will be accusative whether it is transitive or intransitive, we are once again led to the conclusion that this is an instance of unmarked case assignment. One might wonder why, if the complementiser plays no role in case marking the subject, is it obligatory? (18) a * The government were anxious [the bill to be passed quickly]. b * [The bill to be passed quickly] is important for the government. We can think of several possible responses to these observations, which require further investigation than is possible in the present context. The most straightforward of these is to assume that the traditional approach is basically right and that the complementiser does have a licensing role in this construction. However, licensing, being divorced from case assignment, does not affect the analysis of the accusative subject as assigned unmarked case under DCT assumptions. What is missing, however, is an account of what licensing is.

Not so exceptional case marking in English 61

Clearly, proposals which simply adopt something similar to the Case Filter with case factored out of the equation, such as Sigurðoson’s (1991) Proper Head Government Condition, are out of step with minimalist assumptions. But, Marantz’s (1991) original assumption, that DPs are licensed in argument positions through a combination of argument structure considerations and the Extended Projection Principle, is not easily extended to cover the observations we are examining here. As it is not our intention to devise a theory of DP licensing here, we will leave the matter at this point. A second possibility would be to assume that for is not a complementiser in these constructions, but a preposition taking the accusative DP as its complement. If this could be defended, then it might be made to follow from something akin to Marantz’s (1991) suggestion that overt DPs are never licensed in the subject of an infinitive11. The idea would be that although a DP would not be licensed in an infinitival subject position, there would be nothing to stop a PP from appearing in this position. This is probably the least satisfactory account of the obligatory nature of for. There are numerous arguments against taking the for complementiser as a preposition (see, for example, Bresnan 1972 and Huddleston and Pullum 2002), though some of these might be overcome if we supposed a grammatical function for it, similar to the treatment of of within the NP. Nevertheless, the parallels that for shows with other complementisers and the lack of evidence that it forms a constituent with the following DP make for limited prospects as an account for the data in (18). The final approach we will consider is to investigate the conditions on the distribution of overt complementisers. The ungrammaticalities of (18) may well have nothing to do with the licensing of the subject in this position, but rather that the complementiser is required to be overt in its own right. This is suggested by the fact that finite clauses require overt complementisers when in subject position, and this does not seem to be a licensing/case issue: (19) [*(that) the bill was defeated] was a blow to the government’s economic strategy. Things are less clear when we consider clauses in complement positions, however, as here the requirement of an overt complementiser differs for finite and infinitival clauses. Finite clause complements require an overt 11

Marantz suggested that raising to object account for why ECM subjects are licensed. However, we will be taking issue with this account and, following Lasnik (2001) will assume that raising to object is optional and motivated by other considerations when it actually takes place.

Not so exceptional case marking in English 62

complementiser when associated with nominal but not verbal or adjectival predicates: (20) a The fact [*(that) he fell asleep] did not go unnoticed. b We should have known [(that) he wouldn’t be interested]. c It was possible [(that) no one would notice]. With the infinitive, the pattern differs in that the complementiser is obligatory with nouns and adjectives, but not always with verbs: (21) a His demand [*(for) the crowd to disperse quietly] was met with anger. b We didn’t expect [(for) that to happen]. c It was essential [*(for) everyone to remain seated]. Again, however, the problem is the lack of a unified account of the distribution of overt and covert complementisers, without which it is difficult to conclude about the differences between that and for. As with the theory of DP argument licensing, it is not our intention to construct a principled account of complementiser distribution, we just note that such an account may well hold the key to the data in (18) in a way that divorces the complementiser from the accusative subject. 3.2. The accusative subject of ECM constructions Having concluded that the accusative subject in quite a few English constructions should be analysed as an instance of unmarked case assignment, we now turn to consider ECM constructions. One reason to think that these might be different to all the other clauses with accusative subjects is their distribution. Whereas the other constructions are not restricted to a particular position, ECM constructions can only appear in the complement of the ECM verb. Specifically, they are not allowed in subject position: (22)

a b

We believed [him to be unavailable]

[Him to be unavailable] was believed by everyone.

In Standard Case Theory, this observation was accounted for under the assumption that the subject of the infinitive is dependent on the verbal system of the superordinate clause for its case. Later approaches, reintroducing Postal’s (1974) proposals, have argued for a raising to object analysis, in which the infinitive’s subject raises to the object position of the higher clause for case

Not so exceptional case marking in English 63

marking purposes (Pesetsky 1987, Johnson 1991, Lasnik 1995, Bošković 1997). Either way, these assumptions ensure that the infinitival clause be in the complement position of the higher verb. Of course, the same would be true if we divorce case from licensing and maintain that the subject of the infinitive is licensed from the verbal system of the higher clause. From this perspective, this is an issue of licensing, not of case assignment, though it remains to be established what the details of case assignment are in this case: does the subject to object movement create the configuration for accusative case assignment or is accusative case assigned to the infinitival subject before it moves? In this regard, work by Lasnik (2001) and den Dikken (2018) is highly relevant. These both conclude that raising to object is optional and therefore we can find the infinitival subject inside and outside the non-finite clause. Importantly, irrespective of whether movement takes place or not, the subject appears with accusative case, from which we can conclude that the infinitive internal subject position is one to which accusative case is assigned. We will have to go into further theoretical details to determine whether case is assigned independently to both positions or whether it is assigned to the subject position and then carried to the object position when movement takes place. For now, however, we will be content with the conclusion that the subject of the infinitive is a position to which case is assigned. We now face the issue of which case is assigned to the infinitival subject. On the one hand, given that we have determined that all other accusative subjects in English are assigned unmarked case, it seems there is reason enough to assume the same for ECM constructions. On the other hand, there is nothing immediately obviously wrong in assuming that a more standard analysis is correct. From a DCT point of view, this would amount to claiming that the infinitival subject position lies within the same domain as the object of the higher clause. In other words, the infinitival clause does not count as a case domain as far as its subject is concerned. This is usually assumed to be because the clausal complements of exceptional verbs lack a CP, an assumption which is supported by the facts of raising to object: the lack of the phase head (C) renders the clause not a Spell-Out domain and therefore its contents only get spelled out once the next phase head, the v of the superordinate clause, is merged into the structure. If this is so, the subject of the exceptional clause has its case determined under the same conditions as the object, whether or not it has raised to object position. We therefore have two questions to answer: i) does the subject get assigned its case only in the subject position or can it also get case after it moves to object position?; ii) what case is assigned to the subject? These issues are intricately interwoven and not very easy to answer on empirical observation. We will return to them after a discussion of the case assigned to English objects.

Not so exceptional case marking in English 64

4. The accusative case of the object. There is a near universal assumption that English, like all other Germanic languages, is a nominative-accusative language. This is despite the fact that virtually all of its nominal elements display absolutely no case marking whatsoever. The assumption is buoyed up by the observation that most English pronouns (that is, five out of seven) have differing forms in (tensed) subject and object positions. The dearth of case morphemes in English was one of the main reasons for the introduction of the distinction between abstract and morphological case in the standard theory. This has caused numerous problems over the years, not least of which is that it necessitates an account of how the two notions are related, something which has proved to be a rather slippery subject. However, the observation that English makes use of the accusative form of its pronouns to realise unmarked case in the subject positions of virtually all clauses, except for finite ones, raises a number of flags when considering the accusative form of objects. If objects are assigned accusative in the standard way, i.e. as a low dependent case, we have to conclude that English has an accusative realisation of both unmarked and dependent accusative case. While this is not ideal, it would be very difficult to argue that accusative subjects are assigned dependent case as we have pointed out above. It might therefore seem inevitable to assume that English happens to realise different cases in the same way. The consideration of a further issue, however, gives rise to another possibility which might solve the accumulating number of problems. As mentioned earlier, Baker’s (2015) claim that case domains are to be equated with Spell-Out domains means that VP is also a case domain, though for some languages this could be effectively neutralised by the introduction of the notion of a soft domain. Let us suppose that English is not such a language and that the English VP really is a case domain. It would follow from this that objects, being the sole DP in the VP, would be assigned unmarked case. For the majority of nominals this is straightforward as there is generally no case marking on English objects. For the handful of pronouns, however, we would have to claim that they are assigned unmarked case which is realised by the accusative form. But this makes pronoun objects no different to pronouns in most other domains, except for those in finite clauses. From this perspective, the situation is quite straightforward: pronouns with unmarked case have a special form in the finite domain, i.e. what is typically identified as ‘nominative’, and a more general realisation in every other domain, i.e. ‘accusative’. In terms of case assignment, this view claims that English assigns unmarked case to virtually all DPs. The one exception to this, as mentioned earlier, is

Not so exceptional case marking in English 65

genitive, which is the dependent case assigned within the nominal domain. Therefore, it is not totally accurate to say that English has a neutral case system in the way that Chinese has. However, at clause level, there is a neutral system in operation. It is certainly not an accusative language. Within the framework outlined in section 2 of this paper, we claim that English assigns unmarked case to all DPs at clause level and for the majority of them this is straightforwardly reflected in the case forms that realise them. For pronouns, however, there are two case forms which realise unmarked case, which distribute according to the contextual restrictions in (23). The example is for the first person plural pronoun, which has the same pattern as most other pronouns, with some syncretism for a few of them: (23) 1pl. [dependent]  our 1pl. [unmarked]  we / [CFIN – ] 1pl. [unmarked]  us elsewhere There is no need to contextualise the realisation of dependent case, as it is only in the nominal domain that dependent case is assigned. The only contextualisation needed is for the ‘nominative’ form, which is restricted to the finite context. One can see how this situation emerged from what was originally an accusative system, like other Germanic languages. The loss of case morphology in English resulted in a mostly neutral case system. For a small subset of high frequency items, i.e. pronouns, old case forms were maintained as different realisations of the unmarked case in contexts which partially reflected the original system. This would not cause a problem for acquisition, as the vast majority of the data provide evidence for the neutrality of the case system. With this backdrop, it would not be difficult to spot the few lexical elements with contextually determined realisations. Indeed, it would be more problematic to acquire an accusative system, given the data.

5. How exceptional is ECM? From the above discussion, we can see that even if standard assumptions about the nature of ECM are valid and that the case of the exceptional clause’s subject is determined similarly to that of the object, then this is still not very exceptional as objects are assigned the same unmarked case as most other DPs no matter which position they occupy. However, there remains a residue of exceptionality here. Exceptional clauses are generally assumed to lack CPs to allow for raising movements. This assumption might impinge on how case is assigned to the subject of such clauses, even if it does not change which case is assigned. It

Not so exceptional case marking in English 66

would be interesting to explore whether even this amount of exceptionality can be eliminated. The original motivation for assuming the absence of CP in ECM constructions, that government from the superordinate verb would be blocked by the presence of a CP, has no currency in the minimalist approach. However, evidence in favour of raising to object in such constructions provide another argument against the CP status of exceptional clauses. The movement of the subject out of exceptional clauses requires that such clauses not be phases. As C is a phase head, one way that exceptional clauses can be exempt from phase status is for the C to be absent. However, this only holds under one set of assumptions concerning phase theory. The original system of Chomsky (2000) claimed that the complement of a phase head is spelled out once the edge of the phrase projected from the phase head is complete. But Sigurðsson (2002) observed that Icelandic nominative objects accompanying dative subjects indicates that material inside transitive VPs is available at the clause phase. This motivated a different take on the Phase Impenetrability Condition in Chomsky (2001), sometimes referred to as the weak PIC, in which the Spell-Out domain of one phase head is not sent off to the interfaces until the introduction of the next phase head. This keeps open the VP until after the introduction of the next C. Note, however, all things being equal, this should also keep the domain of the complementiser open until the higher v is introduced and thus, in principle, allow for raising to object out of a CP complement. Under these conditions, there would be no need to assume that exceptional clauses are at all exceptional, leastwise as far as case assignment is concerned. If exceptional clauses are CPs, then the contained IP is a case domain in which the subject is determined to be assigned unmarked case. If it remains in situ, and does not raise to object, it will be realised under the relevant contextual conditions, which means that if it is a pronoun is will take the ‘accusative’ form. If it moves, under the claims of Newson and Szécsényi (2022), specifically that CPs are universally hard case domains, then its unmarked case will be determined for it within the CP and carried to the vP internal position. In this position, under the relevant contextual conditions on realisation, the ‘accusative’ form will once again be selected for pronouns.

6. Conclusion In this paper, we have attempted to argue that English exceptional case is far from exceptional in two separate ways. First, it does not involve the assignment of a different case to subjects of ECM constructions to that assigned to any other subject, or indeed most other DPs. The view that English has a neutral case system, at least at clause level, means that all DPs generally are assigned the

Not so exceptional case marking in English 67

same unmarked case. Second, the realisation of the pronoun subject of ECM constructions as ‘accusative’ is in fact unexceptional, as this is the general realisation of unmarked pronoun forms adopted in the language. On the contrary, the nominative realisation of subject pronouns is exceptional from this point of view, as this form is restricted to a very limited context. On the way, we have raised a number of questions which remain to be answered: what is the real nature of DP licensing?; why are complementisers obligatory in certain contexts?; are exceptional clauses really exceptional in any way at all? As none of these impinge directly on the issues of case assignment and case realisation, which has been our main focus, we have not given them the serious consideration that they require. We leave them as open questions for the time being. References Abney, Steven. 1987. The English noun phrase in its sentential aspect. Doctoral Dissertation. MIT. Akmajian, Adrian. 1984. Sentence types and form-function fit. Natural Language and Linguistic Theory 2.1: 1-23. https://doi.org/10.1007/BF00233711 Baker, Mark. 2015. Case: Its Principles and Parameters. Cambridge: Cambridge University Press. https://doi.org/10.1017/CBO9781107295186 Bittner, Maria and Kenneth Hale. 1996. The Structural Determination of Case and Agreement. Linguistic Inquiry 27. 1-68. Bošković, Željko. 1997. The syntax of nonfinite complementation: An economy approach. Cambridge: Cambridge University Press. Bresnan, Joan. 1972. Theory of complementation in English syntax. PhD dissertation. MIT. 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. Chomsky, N. 2001. Derivation by phase. In Michael Kenstowicz (ed.). Ken Hale. A Life in Language. Cambridge, Mass.: MIT Press. 1-52. Dikken, Marcel den. 2018. Secondary predication and the distribution of raising to object. Acta Linguistica Academica 65.1: 87-117. https://doi.org/10.1556/2062.2018.65.1.5 Halle, Maurice. & Alex Marantz. (1993). Distributed morphology and the pieces of inflection. In Kenneth Hale and Samuel J. Keyser (eds.). The View from Building 20. Cambridge, Mass.: MIT Press. 111-176. Huddleston, Rodney and Geoffrey K. Pullum. 2002. The Cambridge grammar of the English language. Cambridge: Cambridge University Press. https://doi.org/10.1017/9781316423530

Not so exceptional case marking in English 68 Johnson, Kyle. 1991. Object positions. Natural Language & Linguistic Theory 9: 577-636. https://doi.org/10.1007/BF00134751 Lasnik, Howard. 1995. Case and Expletives Revisited. Linguistic Inquiry 26: 615-634. Lasnik, Howard. 2001. Subjects, Objects and the EPP. In: William D. Davies and Stanley Dubinsky (eds.). Objects and other subjects; Grammatical junctions, junctional categories, and configurationality. Dordrecht: Kluwer Academic Publishers. 103-12l. https://doi.org/10.1007/978-94-010-0991-1_5 Marantz, Alec. 1991. Case and licensing. In German Westphal, Benjamin Ao and Hee-Rahk Chae (eds.). Proceedings of the 8th Eastern States Conference on Linguistics (ESCOL 8). Ithaca, NY: CLC Publications. 234-253. Newson, Mark. 2018. Default Case. In: Mark Newson and Péter Szigetvári (eds.). The Even Yearbook, 13. Budapest: Eötvös Loránd University 29-55. Newson, Mark. 2019. Unmarked accusative in non-finite domains: the English acc-ing gerund. In: Joseph Emonds, Markéta Janebová and Ludmila Veselovská (eds.). Language Use and Linguistic Structure: proceedings of the Olomouc Linguistics Colloquium 2018, Olomouc Modern Language Series Vol. 7. Oloumuc: Palacký University. 65-78. Newson. Mark and Szécsényi, Krisztina. 2020. Dative as the unmarked unmarked case in Hungarian. In Markéta.Ziková, and Marcin Wągiel (eds.). Linguistica Brunensia 68(2). Brno: Masaryk University. 27-44. https://doi.org/10.5817/LB2020-2-3 Newson, Mark and Krisztina Szécsényi. 2022. Domains of Case Changing and Case Maintaining Movements. In Linguistics Beyond and Within 8. Lublin:John Paul II Catholic University. https://doi.org/10.31743/lingbaw.14963 Pesetsky, David. 1989. Language-particular processes and the Earliness Principle. Paper presented at the 12th GLOW Colloquium, Utrecht. Postal, Paul. 1974. On Raising: One Rule of English Grammar and Its Theoretical Implications. Cambridge, MA: The MIT Press. Schütze, Carson. 2001. On the Nature of Default Case. Syntax 4: 205-238. https://doi.org/10.1111/1467-9612.00044 Sigurðsson, Halldur. A. 1991. Icelandic Case-marked PRO and the licensing of lexical arguments. Natural Language & Linguistic Theory 9: 327-363. https://doi.org/10.1007/BF00134679 Stowell, Tim. 1981. Origins of Phrase Structure. PhD. dissertation. MIT. Cambridge, Mass. Mark Newson Department of English Linguistics, ELTE ORCID 0000-0002-7127-6448 newson.mark@btk.elte.hu

Krisztina Szécsényi Department of English Linguistics, ELTE ORCID 0000-0003-2192-2157 szecsenyi.krisztina@btk.elte.hu