18ne,   PDF
Mark Newson

Default Case*

1. Introduction The notion of default case has been around for quite some time and is probably best associated with the English use of an accusative form where we might expect a nominative one to appear. The obvious case of this is in a fragment pronoun answer to a question which focuses on the subject: (1)

Q A

Who did that? Him.

Later on in this paper, I will argue that the above does not, in fact, exemplify the use of default case, but for now it will do to set the scene. In (1), the pronoun appears apparently out of syntactic context and therefore is not in a structural position in which case could be licensed on it. Yet it appears in a certain form (accusative) and so we must assume that this form is what we get when there is no licensed case, i.e. the default form. However, for many years the notion of default case could not easily be incorporated into the theory for the simple reason that it conflicted with one of the basic tenets of Case Theory: the Case Filter. As default case implies the absence of case and the Case Filter states that all nominal phrases which have no case are ungrammatical, the two notions seriously collide. Not until the Case Filter as a core explanans of nominal distribution was on the decline, starting in the early 1990s, did we find the notion of default case entering into serious theories of case, starting with Marantz (1991) and then with Schütze (2001). The 10 year gap between these two papers is perhaps indicative of the reluctance with which the Case Filter was abandoned and therefore the continuing resistance to incorporating the notion of default into Case Theory proper.

I would like to thank Kriszta Szécsényi for commenting on this paper and Marcel den Dikken for his comments on a previous and somewhat different version of it. The work was originally inspired by a short course given at the Research Institute for Linguistics of the Hungarian Academy of Sciences by Jon Babaljik, which as far as I can recall did not mention default case. It did, however, introduce me to Dependent Case Theory on which the analysis proposed in this paper is based. To the extent that I followed (or properly understood) the advice given by the people mentioned here, then their contribution is gratefully acknowledged. Wherever I deviated from this, I take full responsibility for it.

The Even Yearbook 13 (2018), Department of English Linguistics, Eötvös Loránd University, Budapest ISSN 2061–490X, http://seas3.elte.hu/even. © 2018, Mark Newson

Default Case 30

In this paper I will review Schütze’s by now classic approach to default case and point out what I see as some inconsistencies within it. I will offer an alternative approach based on Baker’s (2015) Dependent Case Theory (DCT).

2. Approaches to default case 2.1. Marantz Marantz’s (1991) paper argued that the licensing of nominal phrases was nothing to do with case, and hence removed the conflict which had prevented default case from being incorporated into the theory. It also outlined a proposal in which default case was incorporated. The general idea was that case is realized on the basis of a disjunctive hierarchy such that lower cases are made unavailable by the availability of ones higher in the hierarchy: (2) -

Marantz’s case realisation disjunctive hierarchy (1991, p.24) lexical governed case “dependent case” (accusative and ergative) unmarked case (environment-sensitive) default case

Lexical governed case (inherent case) overrides all others and hence is immune to interference from grammatical function changing processes, such as passivisation. The novel part of the theory comes through the introduction of dependent and unmarked cases. Dependent case is assigned by a combination of the verb and the inflection either to the object, in which case it is called accusative, or to the subject, in which case it is called ergative. The case is assigned whenever there is a choice of recipient, i.e. in transitive contexts, and so it is dependent not only on the case assigner, but also the presence of another nominal phrase. Unmarked case is assigned to nominal phrases which are not assigned dependent case, i.e. the subject of a transitive verb with an accusative object, the object of a transitive verb with an ergative subject and the subject of an intransitive verb. Thus when the object is assigned dependent case, subjects are assigned unmarked case, in which case it is typically referred to as nominative. When the subject of the transitive is assigned dependent case, the unmarked case is assigned to the object and the subject of the intransitive verb and this case is often called absolutive. Marantz does not give a detailed account of default case. Its final position in the disjunctive hierarchy indicates that it is to be assigned to any nominal phrase which does not get any of the other cases. It is unclear, however, how this is supposed to work given that unmarked case assignment seems to be the

Default Case 31

‘elsewhere’ condition, assigned to nominal phrases which do not get either lexical governed or dependent cases. Thus, there appears to be no room in the theory for the default. It is possible that this problem might be approached through the ‘environment sensitivity’ of unmarked case, by which it is meant that there may be different unmarked cases for certain structural environments: nominative for clauses and genitive for nominal contexts, for example. Perhaps there are environments in which there is no unmarked case and so default is able to emerge in these. However, this is not an approach Marantz explicitly adopts. 2.2. McFadden McFadden (2007) develops Marantz’s (1991) proposals, providing a feature based analysis in which accusative is defined by the assignment of a feature [+inferior], which is assigned in the presence of a c-commanding DP1. McFadden seeks to address the apparent conflict in Marantz’s system between unmarked and default case by collapsing the two into one. In this system, nominative is unmarked because it is not assigned a case feature. It will therefore emerge in environments where no case features are assigned. The subject position is one environment where no case feature is assigned, but there are others which are typically thought of as involving default case. He supports this view by the fact that nominative is the default case in many languages. For example, it is the case of many left dislocated nominal phrases which show no concord with the resumptive in a structural position (McFadden 2007, p. 229): (3) a

b

c

d

1

Der Hans, mit dem spreche ich nicht mehr the NOM Hans with him DAT speak I not more ‘Hans, I don’t speak with him anymore.’ (German) Vanjai, ego ja ne ljublju John NOM him ACC I don’t like ‘John, I don’t like him.’ (Russian) al-kitaab-u qara?t-u-hu. the-book-NOM read-lSG-it ‘The book, I read it.’ (Arabic) Strákarnir við þá hafði aldrei verið talað. boys-the NOM with them ACC had never been spoken ‘The boys, they had never been spoken with.’ (Icelandic)

It is further stipulated that the c-commanding DP should not bear a non-structural case to account for why accusative is not assigned to objects in quirky subject constructions.

Default Case 32

To account for why English uses the accusative form for left dislocated DPs and other cases where the default appears, McFadden claims that it has a different case system in which the accusative is unmarked and the nominative marked. Here he has to diverge from Marantz, as he does not want to say that English is an ergative language, assigning the dependent case to the higher DP. Instead, McFadden claims that nominative case in English is a marked case assigned by finite inflection to the subject under a strict locality condition. Accusative is therefore unmarked, having fewer assigned features and appearing wherever nominative is not assigned. McFadden’s account lacks the details that would enable a more thorough evaluation. He claims that nominative is generally unmarked because it lacks the [+inferior] feature assigned to the object and therefore has fewer features assigned to it. However, how the assignment of nominative case in English and the assignment of the [+inferior] feature to the object are related is unclear. If the object is assigned [+inferior] when the subject is assigned nominative, then both have the same number of features and so one is no more marked than the other. Thus either the nominative feature must count more in calculating markedness or the assignment of the [+inferior] feature must be blocked. But how this all works out is not discussed by McFadden. While I will agree with McFadden that English differs from other languages in that it has accusative case as unmarked and default, I disagree that these two usages can be collapsed into one. I will demonstrate later that there are differences in the behaviour of unmarked and default accusative that would not be predicted if the two are simply taken to be the emergence of the unmarked. 2.3. Schütze Following Marantz (1991), Schütze (2001) assumes that the Case Filter is not responsible for the distribution of nominal phrases and therefore that ‘licensed’ DPs can emerge from the grammar without being assigned case. Adopting the ‘late insertion’ assumption of Distributed Morphology (Halle and Marantz, 1993), he claims that default case is simply the form chosen to spell out such a DP. From this perspective, default case is not assigned at all, either as a marked or unmarked case, but is the result of the absence of case. In his 2001 paper, Schütze identified eight contexts in English in which he claimed the default can be found. However, in his 1997 dissertation the list extended to 12: (4)

a Madmag sentences: What? Her cheat on you!? Never! b Argument small clause: Her in New York is what we must avoid. c Adjunct small clause: Him tired, they decided to camp for the night.

Default Case 33

d Acc-ing gerund: Him liking beans surprised them. e Predicate nominal: It was us. f i) Left dislocation: Me, I like beans. ii) Appositive: I think the best student, her, should be president. g Ellipsis: Me too! h Subject of understood predicate: Everyone but them gets on John’s nerves. i Pronoun fragment answers: Who did it? – Me. j Gapping: We can’t eat caviar and him beans. k Conjoined subject: Us and them are going to rumble. l Modified pronoun: The real me is finally emerging. The reason why Schütze takes these contexts to be ones involving default case is that none of the (bold) accusative pronouns in (4) can be said to be associated with a verb or preposition that could assign accusative case to it. But this is a very theory dependent position to take and Schütze adopts a rather traditional theory of case assignment. Despite following Marantz’s argument that licensing nominal distribution is not carried out by case, which was one of the bedrock assumptions of standard case theory, Schütze does not adopt Marantz’s version of case theory. Clearly, what assumptions one makes about the conditions of case assignment will affect claims about what positions case is and isn’t assigned to and therefore the whole issue of whether or not we take the appearance of an accusative nominal phrase to be an instance of default rests entirely on our theoretical assumptions. This presents us with a rather difficult situation. If our theory decides when case is assigned and when it is not, i.e. when the default emerges, how can we determine the accuracy of the theory? There are no theory independent observations to enable us to do this. Fortunately, I believe there is a way out of this dilemma as there are independent observations concerning the behaviour of default case which enable us to see a theory independent empirical difference between assigned accusative and default accusative. We will discuss these observations in the following section.

3. Hypercorrective nominative and default case Let us first consider what is not a good way to distinguish between assigned and default accusative. The example that Schütze gives for an argument small clause (4b) involves such a construction in subject position and not in a VP internal position, such as in (5): (5)

I consider [her smart]

Default Case 34

Perhaps the reason for this is that Schütze assumes, as is standard, that such constructions involve an ECM type analysis. A standard argument in support of this analysis is that, when the governing verb is passivised, the subject of the small clause must move to the higher subject position: (6)

a * b

it was considered [her smart] she1 was considered [ t1 smart]

But under the assumption that case does not play a role in licensing nominal distribution these observations no longer necessarily lead us to assume that the accusative case on the small clause subject in (5) comes from the governing verb. The reason why the subject must move in (6) has to do with licensing, but not necessarily case. Therefore, it could be maintained, against standard assumptions, that the accusative case on the small clause subject, no matter whether the construction is VP internal or external, is an instance of default case. Indeed, if we took this argument to its logical conclusion, we might as well claim that all instances of accusative are default, even those in object positions, as there is no evidence that the verb assigns case to it, only that this is a licensed position that happens to show accusative case. This is close to the position that McFadden was advocating, though perhaps even more extreme. Is it possible, under the assumption that case assignment and nominal licensing are distinct, to distinguish between assigned case and default? An observation emerging from Schütze’s data is that sometimes the accusative pronoun can be exchanged for a nominative one and sometimes this is impossible. To demonstrate, note the clear distinction between the following: (7)

a b

Him/*he liking beans surprised them. why couldn’t he take my car and me/I his2

Although Schütze (2001) notes these differences, he has very little to say about them. The nominative forms, where they are allowed, he attributes to hypercorrection: a stylistic variant influenced by the over application of a spurious or otherwise grammatical rule3. He points out that as default case is a 2

3

Schütze (2001) assigns a question mark to the nominative form in this example, indicating that he finds is somewhat odd. I find it acceptable as a hypercorrective form. Perhaps it is this status that Schütze’s question mark is signalling. For example, the claim that it is ungrammatical to ‘split’ an infinitive in English (to boldly go) is based on Latin, which had a morphological infinitive. Clearly the over extension of this ‘rule’ to English is totally spurious. However, the claim that all nominal phrases associated with the subject of finite clause is based on a non-spurious observation that such subjects are typically nominative. However, this ‘rule’ is often over applied to instances

Default Case 35

marginal grammatical notion, not something assigned by a principle of the grammar but something which emerges out of the failure of such principles applying, we might expect that it will be susceptible to interference from considerations which lie outside the linguistic system, such as hypercorrection. This seems reasonable speculation. But if this is what is going on, why are only some of the environments that Schütze identifies as involving the default affected while others resist the hypercorrective nominative? I suggest that the difference between the two types of context is simply that the first really are cases of the default and the others are not. My argument is the same as Schütze’s: the default is marginal and therefore susceptible to outside interference. Assigned accusative is part of the grammatical system and therefore is harder to influence from considerations external to the linguistic system. For example, we do not get hypercorrective nominative with objects and subjects in ECM contexts precisely because these are cases of assigned accusative4: (8)

a * b *

Mary saw he. I believe he to be smart.

I contend therefore that those cases in Schütze’s list in which the nominative is resisted involve assigned accusatives rather than the default. Clearly Schütze’s case theory does not predict this division. In his theory all the environments in the list are ones in which case is not assigned. I argue therefore that Schütze has not adopted the correct theory of case assignment. In the following section I will introduce a theory of Case assignment which will be shown to make distinction between case assigning and default environments more accurately.

4. Dependent case theory à la Baker Before presenting the version of dependent case theory as proposed by Baker (2015) and showing how this helps to make the right division between case assigning and default domains, let us first return to Marantz’s (1991) version of the theory. In this, dependent (accusative) case is assigned to a nominal phrase which is lower in a structure than another nominal phrase. Nominal phrases which are not in such a configuration get unmarked case.

4

where the grammatical system does not determine the presence of a nominative, as in the case of (7b) Having said this, the real enthusiast can infiltrate the grammatical system if they really try. I was once told a story by Eric Fudge of a parish vicar who in his sermon offered the following observation: “God comes to you and I. He comes to you, and he comes to I”.

Default Case 36

Putting aside the question of how, under these assumptions, default case could ever emerge, if we apply these principles to the structures in Schütze’s list, we do not get much different a result to Schütze’s. Five of the 12 environments in the list involve a nominal phrase in a subject position and most of the others have the nominal phrase in a position for which there is no superior nominal phrase present. Thus, from this perspective, we would seem to have to assume that none of the environments involve an assigned dependent case and given that the form of these phrases is accusative we would be forced to agree with Schütze that they are instances of the default. Although Baker (2015) does not discuss default case assignment in any detail, he does provide a more fleshed out theory than Marantz’s. The basic assumptions are essentially the same, but Baker places more emphasis on the notion of a domain: the part of the structure in which the principles of case assignment operate. The domain has two basic functions in Baker’s theory. First it delimits the application of case assignment rules to given areas of a structure. Thus, we wouldn’t expect the appearance of an object in a subordinate clause to have any effect on the case assignment to a subject in a superordinate clause. Therefore the assignment of dependent case can only take into consideration the nominal phrases present in a limited part of the structure rather than the structure as a whole. Baker assumes that domains equate to phases and therefore the principles of case assignment apply within phase boundaries. The second function of a domain is to provide a structural area in which a language can define different dependent and unmarked cases. This is the ‘environment-sensitivity’ that Marantz (1991) referred to when he claimd that nominative can be the unmarked case in a clause and genitive in a DP. Baker extends this with the notion of a domain to account for the assignment of structural dative as a dependent case within the VP domain, for example. Newson (forthcoming) argues that English selects a standard nominative case as the unmarked case in finite clauses, but takes accusative to be unmarked in all other domains. This accounts for the widespread appearance of the accusative in English, from the subject of the acc-ing gerund to the the goal in the double object construction. The idea links to McFadden’s (2007) claim that accusative is unmarked in English. But instead of claiming that nominative is therefore marked, it is assumed that both cases are unmarked in different domains. It is the use of domains which will help us to determine the difference between assigned (unmarked) accusative and default accusative in what follows. Something that has not been picked up on in previous work is a potential third function of a domain which directly impinges on the theory of default case. As we have pointed out, dependent case theory faces a problem with

Default Case 37

respect to default case. If unmarked case is assigned whenever dependent case is not, there doesn’t appear to be any room in the theory for an unassigned default. However, once the notion of a domain is introduced, we can utilise this to ensure that there will be instances in which case will fail to be assigned. This follows straightforwardly if we make the principles of case assignment contingent on the presence of a domain, which seems reasonable given the central position of domains in case assignment. Without a domain, dependent and unmarked cases are undefined and without a domain the theory does not know which nominal phrases have to be taken into consideration to determine what to assign case to. Thus I propose a slight modification to Baker’s principles of case assignment5: (9)

Given a domain D, i) if there are two nominal phrases in D such that one ccommands the other, then assign dependentD case to: a) the c-commanded one (accusative) b) the c-commanding one (ergative) ii) assign all other nominal phrases in D unmarkedD case.

This theory ensures that assigned dependent and unmarked cases will only appear in the context of a domain and that in cases where no domain is defined, no case will be assigned. In the latter circumstances, any nominal phrase present will be spelled out with the default form. An important point needs to be made before we continue. It is not just that the constituent containing a nominal phrase is not defined as a domain that renders the principles of case assignment inapplicable. Clearly, not all constituents are possible domains. Yet a domain can still be established for a nominal phrase no matter how many non-domains it is contained in. Recall that Baker equates domains with phases. This means that every domain must be a phase. However, it does not necessarily mean that every phase is a domain. Specifically, I will assume that if a phase cannot be defined as a domain, the nominal phrases contained in it will not have a domain and therefore cannot be assigned dependent or unmarked case. It is exactly here that the default will emerge.

5

In (0), dependentD and unmarkedD refer to the dependent and unmarked cases, respectively, defined with respect to domain D.

Default Case 38

5. Testing the theory In this section we will return to Schütze’s (1997) ‘default contexts’ and examine them to see if they show signs of the presence of actual default. Against this, we will check the predictions of the case theory outlined in the previous section to see whether it is a more accurate predictor of default contexts than Schütze’s theory. 5.1. Coordinated pronouns I will start the discussion with the case of coordinated pronouns as these raise a number of issues that impinge on the present discussion. The examples of this construction that Schütze discusses are labelled as ungrammatical with nominative pronouns: (10)

a * b * c *

We and they are going to rumble. She and we have been friends for ages. Did your parents or he pick up Mary

However he dismisses examples involving first person singular pronouns, such as (11), as “prescriptive dogma” (1997, p. 59, fn. 73): (11)

You and I need to talk.

By ‘prescriptive dogma’ I assume he means what I have been calling hypercorrection and therefore he seems to agree with me that this is a context in which external (prescriptive) interference is possible. A number of points need to be raised however. It would seem from the fact that Schütze singles out examples such as (11) as a special case, claiming them to be not part of any grammar of English, that he assumes the possible appearance of nominative forms in other ‘default contexts’ are instances of grammatical constructions. However, there are similarities between the behaviour of nominative pronouns in these contexts, such as number and person restrictions placed on them, which would indicate that they are instances of the same phenomenon. Thus, if coordination provides a context in which hypercorrection is possible, then so do the other contexts which allow the appearance of a nominative pronoun to one degree or another. For me, the variable acceptance of a nominative form is precisely the observation which allows us to accept these contexts as involving the default. A second point concerns Schütze’s assumption that hypercorrection in these cases only involves the first person singular. This is not true. The

Default Case 39

following seem to me to be perfectly possible cases of hypercorrection, restricted to a formal and slightly pompous register: (12)

a b c

He and his sons had been painting the house all day. They and their ilk will not be remembered by history. We and the other members of the board agree on this point.

I agree with Schütze’s judgements that there are cases where the nominative is harder to accept: (13)

a ? We and they are in communication over the matter. b ?? She and we have been friends for ages. c ?? I and you need to talk.

As a comparison between (11) and (13c) indicates, the conditions on the acceptability of hypercorrection are rather varied; it clearly is not just a matter of person and number but also one of conjunct ordering. The same is not true for the default: (14)

a b

You and me need to talk. Me and you need to talk.

I suggest that it is exactly these odd restrictions on the use of the nominative that indicate its hypercorrective status. The restriction to putting the first person nominative second can probably be put down to an issue of politeness in the sense that it is impolite to put yourself in front of others (see Parrott 2009 for a similar view). Inasmuch as the appearances of nominative pronouns in other ‘default contexts’ show similar odd restrictions we should take these to have the same hypercorrective status.6 Having established that coordination yields a default context, the next question to deal with is: why? From the perspective of the theory developed above, we must conclude that coordinations do not count as domains (though they are phases) and hence the conditions for case assignment are not met 6

Krisztina Szécsényi wonders if there might be a restriction placing pronouns before lexical DP operating in cases such as (12). This seems to be accurate inasmuch as these examples would be worse with the pronoun second. However, the first person last restriction seems to override this: i) Your mother and I will never see eye to eye. I assume that both of these restrictions are based on prescription as neither seem to operate with accusative pronouns: ii) a Me and your mum aren’t talking. b Your dad and her don’t get on.

Default Case 40

within them. Baker (2015) proposes that c-command is a central notion in dependent case theory inasmuch as the nominal phrases which trigger the assignment of dependent case must be in a c-command relation with the other nominal phrase. Thus we might assume that if c-command relationships cannot be established a domain cannot be properly formed. Although the traditional distinction between coordination and subordination holds that the latter is hierarchical while the former is not, a more recent trend is to analyse coordination as involving an X-bar structure, where the first conjunct c-commands the second (Munn, 1987; Zoener 1995; Johannesson, 1997). However, Progovac (1998) demonstrates that the empirical evidence supporting the claim that the first conjunct c-commands the second does not stand up to scrutiny and, if anything, the data actually support the claim that there is no c-command relation between conjuncts. For example, the following are often cited as supporting the assumption of hierarchical structure in the coordination construction: (15)

a b *

John1’s dog and him1 went for a walk Him1 and John1’s dog went for a walk

The ungrammaticality of (15b) follows if the r-expression in the second conjunct is bound, and therefore c-commanded, by the pronoun in the first. However, Progovac points out that the same observations hold across independent sentences where c-command relationships cannot be defined: (16) * He1 finally arrived. John1’s dog went for a walk. This casts doubt on whether the observations in (15) have anything to say about the structural relation between the two conjuncts. Another observation which is often enlisted to demonstrate hierarchical structure in coordination concerns the possibility of a bound variable pronoun in the second conjunct: (17)

Every man1 and his1 dog went to mow a meadow.

It is well established that pronouns must be c-commanded by a quantified expression if they are to act as a bound variable. However, Progovac counters that due to quantifier raising the pronoun would be c-commanded by the quantified phrase at LF even if it is not prior to the movement. Therefore data such as (17) is inconclusive about whether there is a c-command relationship between the first and second conjunct.

Default Case 41

On the other hand there are data which point to the conclusion that there is no c-command relationship between the two conjuncts. For example, the first conjunct cannot be the antecedent of a reflexive within the second and a negative element in first conjunct position does not licence a negative polarity item in the second: (18)

a * b *

Either John1 or a picture of himself1 will suffice. He chased nobody and any dog.

If there was a c-command relationship established between the first and second conjunct, both of these sentences should be grammatical. Their ungrammaticality points to the opposite conclusion. Whatever the reason that a c-command relationship cannot be established within a coordinate structure, be it to do with the structural organisation of these constructs or something else, I contend that this is enough to disqualify the construction as a domain, as domains must be such that c-command relationships are possible to define within them. Therefore, as a domain cannot be established here, case is not able to be assigned and the default accusative emerges. 5.2. Left dislocation Left dislocation is context where the nominative form is resisted and therefore shows signs of assigned accusative rather than the default: (19)

a b c

Me/*I, I like beans. Him/*he, he doesn’t know anything. Us/*we, we’re in this together.

The left dislocated phrase sits outside the main IP and therefore is external to this spell-out domain. Thus, after the main IP is spelled out, there must be one more phase, call it the final phase, in which the elements in the left periphery are spelled out. Therefore the left dislocated element is included in the final phase. There is no reason to believe that the final phase could not count as a domain. The fact that c-command relationships can be established between a left dislocated element and elements within the clause, suggest that it is a normal part of the structure: (20)

a b

John1, he1 likes beans.

• Him1, John1 likes beans.

Default Case 42

We can therefore establish the following three points: i) The left dislocated element is within a domain; ii) The left dislocated element is the only nominal phrase in this domain; iii) This domain is distinct from the finite clause. It follows from i) that the left dislocated element will be assigned a case. It follows from ii) that this case will be unmarked and it follows from iii) that the unmarked case will be accusative, as this is the unmarked case for all domains except for the finite clause. 5.3. Pronoun fragment answers Next, consider the ‘default context’ which we discussed at the beginning of this paper: a pronoun answer to a question. This is the archetypical instance of default case because apparently the pronouns appear outside of any observable structural context. Yet Schütze judges that the nominative form is not possible here. I agree with this judgement and the fact that it seems consistent over the range of person and number variants suggest that, counter to expectation, this is not an instance of the default: (21)

Who wants ice cream?

• I/*he/*we/*them/etc.

A common explanation for the use of the accusative pronoun to answer subject focussed questions is that such answers derive from heavily elided cleft constructions: (22)

It is me that wants ice cream.

As Schütze points out, this is hardly credible. For one thing, the pronoun answer does not carry the identificational focus interpretation that the pronoun in cleft position does (É. Kiss 1998): a simple answer ‘me’ does not entail that I am the only person that the answer is relevant to. However, just because (22) is probably not the underlying structure from which the pronoun answer is derived does not mean to say that there could be no other possible underlying structure. One only has to look at what happens in similar situations in other languages to see that we do have to posit some underlying structure. For example, in Hungarian a simple pronoun answer to a question takes the case relevant for the focus of the question: nominative for a subject and accusative for an object:

Default Case 43

(23)

a

b

Ki szereti a kàposztàt? who likes the cabbage “Who likes cabbage?” Ki-t választ-ot-tak who-ACC chose-past-2.pl. “Who did you choose?”

Ő. he “Him.” Ő-t he-ACC “Him.”

Without assuming some underlying structure to provide the context for the pronoun answers in (23), it is difficult to see how the relevant case could be determined. If we posit an underlying structure for pronoun fragments in Hungarian, it would be an odd claim that English lacks such underlying structures. This does not seem a natural parameter of linguistic variation and it would be extremely hard to explain how such a difference could be acquired by language learners. Furthermore, it is not the case that all English pronoun fragments are accusative. A genitive pronoun is appropriate to a question which focuses on the possessor: (24)

Whose is this?

Mine/*me.7

Again, we are forced to posit some underlying structure to account for the particular use of the genitive in this case and therefore we have direct evidence that English does make use of such structures. Presumably, this underlying structure is clausal and therefore it is reasonable to assume that it counts as a domain. This is enough to discount the use of the default, and so half the story has been told already. The other half of the story has to do with why English assigns the accusative to both the subject and the object of this structure, while Hungarian assigns nominative to the subject and accusative to the object. The accusative case assigned to the object in both languages is not problematic: it is an instance of dependent case assigned to the lower of the two nominal phrases in the domain. That English assigns unmarked accusative to the subject indicates that whatever the status 7

Krisztina Szécsényi points out that the genitive form my would not be possible here and wonders if this might have some connection with why the accusative but not the nominative is available in non-possessive contexts. I fail to see the connection, however, between the possessive pronoun and the nominative pronoun, on the one hand, and the possessive determiner and the accusative pronoun, on the other. It seems to be that a better explanation is simply that the possessive determiner cannot be used by itself in the same way that articles cannot. Thus a better parallel is drawn between articles and the possessive determiner and demonstratives and the possessive pronoun: i) a Whose is this car? Mine./*My. b Which is your car? That./*The.

Default Case 44

of the elided structure, it does not count as a finite clause. The English finite clause has nominative as its unmarked case, while all other domains have accusative as unmarked. The Hungarian data does not necessarily contradict this assumption as it is probably the case that Hungarian has nominative as the general unmarked case, having no obvious instances of an unmarked accusative. The unmarked possessor in Hungarian is nominative, showing that nominative is unmarked in the DP domain. Thus we can assume that the elided clauses have the same non-finite status in both English and Hungarian and that the different cases assigned to the subject result from different choices of unmarked case for this domain. The final part of the story has to do with why the elided domain is not a finite clause. This is a much more difficult question to answer as there is very little evidence, apart from the case assigned to the subject of the construction, concerning the nature of this construction. I can do no more than speculate at this point as to why the construction is not a finite clause. I will therefore put the issue to one side for future consideration. Schütze’s cases of ellipsis which involve the use of a pronoun and some clausal modifying element but no inflection can be handled in much the same way as fragment answers: (25)

a b

I want a go. I didn’t see anything.

• I/*he/*we/*they too!
• I/*he/*we/*they neither.

It is reasonable to assume that there is some underlying structure for these cases of ellipsis and that these are clausal in nature. Therefore, as with fragments, we would expect them to count as domains and the default is not possible. 5.4. Mad Mag sentences, small clauses and gerunds These are perhaps the most straightforward of Schütze’s ‘default contexts’ and I will not spend much time discussing them. They all involve an obvious clause-like structure and none of them allows the use of any nominative pronoun in subject position: (26)

a b c d

• He/*I/*they/*we wear a onesei?!
• She/*I/*they/*we in New York is what you want to avoid.
• He/*I/*they/*we tired, we decided to camp for the night.
• He/*I/*they/*we liking beans surprised them.

It is easy to see that these contexts are all good candidates for domain status and hence it is unsurprising that they do not show signs of default case.

Default Case 45

There is one observation that stands some discussion. One kind of gerundlike structure does allow a nominative subject. This is sometimes referred to as the nominative-absolute construction: (27)

Mike expected to win the game, he/him being the best athlete in the school.

The fact that there is a possible alternation between the nominative and accusative subject in this construction appears to indicate that this is an instance of the default. However, there are a number of observations which suggest that appearances might be deceptive in this instance. There are a number of limitations on the construction which indicate that it is not a simple gerund. First, its distribution is limited to adjunct positions and it never appears in argument positions. Second, the verbs that appear in this construction are limited to being or having. A search of the British National Corpus revealed that of all the gerund-like constructions with a nominative subject, all 73 were absolute clauses, of which 49 had being and 24 had having. No other kinds were present. This seems to indicate that these constructions are fossilised and are not available for grammatical manipulation. Therefore the use of the nominative is not an instance of hypercorrection, but a reflection of a historical form. Indeed, when one looks at the history of these constructions the fossilisation process is quite clear. As reported by He and Wu (2015), absolute clauses developed from an old English construction in which a dative subject was used. The nominative subject first appeared around the middle of the 14th century, though it did not really take off until the beginning of the 1800s and its use stabilised fairly soon after this. It wasn’t until the 1920s, however, that the accusative subject made an appearance in the construction and this was followed by rapid growth in popularity and the accusative usage overtook the nominative during the 20th century. As a historic fossil, nominative absolute clauses cause no problems for the present analysis. 5.5. Modified Pronouns Schütze’s (2001) observations concerning modified pronouns offer an interesting case as there appear to be different conditions applying in different cases. For pronouns which are premodified there seems to be no possibility for

Default Case 46

the use of the nominative and hence these do not appear to be instances of the default8: (28)

a b

The real me/*I/him/*he is finally emerging. Lucky/poor me/*I/him/*he …

On the other hand, with certain postmodification a nominal variant is possible: (29)

a b

We/us three have to be leaving now. We/us linguists are a crazy bunch.

Note that this usage shows the standard signs of the default in that there are differences depending on number and person: (30)

a

Them/*they physicists should know better.

However the construction also shows stricter limitations in that in some cases neither the nominative nor the accusative pronoun is possible: (31)

a b

• I/*me linguist

You/*he/*him fool!

At the very least what can be said is that the cases of premodified pronouns are very different from those in (29)-(31). The former seem to be cases of assigned accusative, not default. They involve instances of pronouns embedded within a larger nominal structure and as DPs are domains in other cases, I see no reason why they wouldn’t be here. It is harder to know what to say about the other examples.9 The fact that this construction seems to be limited to a subset of the pronouns, both nominative and accusative (we, us, you, them) seems to suggest that these are more like set phrases (like “dear me”, see footnote 6). If this is so, they are not relevant to the present discussion. 5.6. Subject of understood predicate Some cases that Schütze (1997) refers to as subjects of understood predicates seem to fall into the coordination category: 8

9

He includes the expression “dear me” in this category, though this is obviously a set phrase as there are no other variants of it (*dear you/*dear him/*dear us/etc.) and therefore uninformative to the discussion. Krisztina Szécsényi suggests that it is the plural pronouns which make up this group. However, the use of you in this way is not restricted to the plural (‘you fool/fools).

Default Case 47

(32)

Everyone but them/*they gets on John’s nerves

It seems that he assumes that the nominative is impossible in these contexts, but this is untrue: (33)

No one but I/we/?they/?he can save the world.

Again we see odd discrepancies in the use of different nominative pronouns, which as mentioned is indicative of a default context. As cases of coordination, these examples are accounted for in the same way as discussed in the previous section. Other examples he gives for this environment are different as the nominative really is impossible: (34)

a b

Students smarter than her/*she get no scholarship. The Jets did that, not us/*we.

In his (2001) paper, Schütze claims examples such as (34a) to be irrelevant as the comparative could be analysed as a preposition which might be responsible for the accusative case. From the current perspective, an equivalent position can be adopted. Assuming the pronoun to be inside a PP, its domain will be the clause that contains it and under standard assumptions it will be assigned dependent accusative case, as it is c-commanded by the subject in the same domain. Therefore we do not expect it to alternate with the nominative as it is not an instance of the default. (34b), on the other hand, would seem to be better treated as a case of ellipsis. 5.7. Predicate nominal The data that Schütze (1997) presents indicate that predicate nominals cannot appear in the nominative: (35)

a b c

It was us/*we. There’s me/*I. The murderer is her/*she.

However, some extremely common hypercorrections are being ignored here: (36)

a b c

It is I. This is he. I am he as you are he as you are me and we are all together.

Default Case 48

Presumably Schütze would react to these the same way as he did to nominative pronouns in coordinations and claim them to be not part of proper grammar. As I discussed in the section on coordination, on one level I agree with this, precisely because these are instances of hypercorrective interferences from outside of the linguistic system. But that is precisely what shows these instances of the accusative in these constructions to be default as they are susceptible to such interference. The question of why the default emerges in this position is a little more tricky to answer. It is probably nothing to do with the availability of a well defined domain in this case as there is no reason to believe that there is anything structurally unusual about the construction. Probably then it is the predicate status of the nominal that is responsible for the lack of assigned case. I assume that the case system only sees arguments, or DPs which are predicated of in cases like left dislocation, and that predicates are looked over. Yet they must be realised by some form and it is precisely those pronouns which lack case features which are realised by the default. 5.8. Gapping I have left the gapping data till last as they are the most problematic for the present theory. The data presented by Schütze (2001) shows possible alternation between accusative and nominative subjects in the gapped clause: (37)

a b c d e f

We can’t eat caviar and him/*he beans. For Mary to be the winner and us/??we the losers is unfair. Why couldn’t he take my car and me/?I his. None of the kids will eat anything, except them/*they the crisps. What did everyone eat? Me/*I beans; him/*he rice and them/*they carrots. She grew up in Jacksonville, me/??I in Tallahassee.

There are two ways we might approach these observations. One would be to claim that in this case this alternation is not a matter of the default form being replaced by a hypercorrect one, but that an ambiguity in the structure allows both assigned nominative and accusative forms. The other would be to claim that there is something about the gapped construction which discounts it as a domain and therefore the default can emerge which can subsequently be hypercorrected. I will consider these options separately.

Default Case 49

Currently, the most widely accepted analysis of gapping assumes that it involves the coordination of VPs, the second of which undergoes VPdeletion10 (Siegel, 1987; Coppock 2001): (38)

IP DP John1

IP I

VP

can’t

VP t1

and VP

V

DP

eat caviar

VP DP Bill

VP VP

DP

eat t2 beans2

Johnson (2014) claims that this is the best analysis around at present, addressing certain properties of the structure which have been problematic for other analyses. However, he also points out that it is not perfect. For example, while (38) provides an adequate analysis for the interpretation involving the modal scoping over both conjuncts, it does not do so well with the interpretation where the modal scopes over the conjuncts individually : (39)

a b

it can’t be the case that John eats caviar and Bill eats beans John can’t eat caviar and Bill can’t eat beans

The second interpretation seems to require that the gapping structure involves conjunction at the clausal level. Other observations also point to a possible clausal conjunction analysis. For example, Johnson (2014) argues that Pesetsky’s (1982) observation that gapping can involve coordinated whquestions can be accommodated in a VP conjunction analysis if we assume that the second VP has an in situ wh-phrase:

10

The following diagram is based on ones given in Johnson (2014) which simplify certain structural issues, such as the configuration of the conjunction structure. As these issues will not impinge on the present discussion, the representation given here will suffice.

Default Case 50

(40)

Bill asked [CP which books1 I3 gave2 a b * c *

[VP [VP t3 t2 t1 to Mary] and [VP I gave which books to John]]] [VP [VP t3 t2 Mary t1] and [VP I gave which books John]]] [VP [VP t3 t2 Mary t1] and [VP I gave John which books]]]

Johnson supports the VP conjunction analysis with the observation that it provides a straightforward account for why (40b) is ungrammatical as the double object construction involves the goal preceding the theme. However, he fails to mention that the analysis predicts the grammaticality of (40c) where the word order is expected, but the in situ wh-phrase is at the back of the clause. The fact that this is ungrammatical undermines Johnson’s claim that the second conjunct involves wh in situ. But if there is wh-movement in (40a) this would be evidence that the construction involves conjoined CPs. The same conclusion can be reached considering the following observations: (41)

Bill, I discussed the plans with and Fred, the costs

This example demonstrates that we can get topicalisation in the second conjunct as the order of the constituents is not one we would expect without movement. Topicalisation is usually seen as a high movement, certainly higher than the VP. Again, we are led to accept that a conjoined CP analysis is at least a possible one for gapping structures. Nothing that I have discussed above denies the possibility of VP conjunction in gapping, and so it might be that there are two possible analyses which coexist side by side. This, then, might give rise to different case realisations for the subject in the gapped constituent. If CPs are conjoined we would expect that the subject would appear as nominative. If VPs are conjoined, then we might claim that the VP internal subject would appear with accusative case. This is the analysis proposed by Murphy (2017), who argues against Schütze’s claim that gapping involves the default. Adopting DCT, Murphy claims that the raised subject of the first conjunct interacts with the VP internal subject of the second conjunct, determining that dependent case be assigned to the latter. However, if the nominative subject were the result of a possible clausal coordination analysis of gapping we would expect that this form would have the same status as the accusative. But Schütze’s grammaticality judgements in (37) clearly show that this is not so. Nominative subjects are less grammatical and show the kind of variation we might expect of hypercorrection rather than a grammatically assigned case. Moreover, under this analysis we would

Default Case 51

predict a strong correlation between phenomena that indicate clausal conjunction and the use of the nominative in the second conjunct: (42)

a b

You can’t eat caviar and me beans.

it can’t be the case that you eat caviar and I eat beans

you can’t eat caviar and I can’t eat beans

it can’t be the case that you eat caviar and I eat beans

you can’t eat caviar and I can’t eat beans