2.1 The effect of negation

In order to see more clearly the discourse functions of Scandinavian clause-final subjects, we can make use of the fact that negation, like other operators, targets the focus of the clause. Consider the Norwegian examples in (7).

1. (7)

In all three examples in (7) it is the negation of the accentuated constituent that distinguishes the assertion from the presupposition. That is, the focus is the accented constituent with the negation applied to it.

Now consider the example in (8a), which is a negated clause with a clause-final subject. The most natural reading is one where the negation applies to the subject, so that the negated subject represents the focus of the clause.

1. (8)

It is however possible to force other constituents to be interpreted as focus. In (8b), accentuation falls within the locational phrase, and accordingly, the focus is taken to be the negation of this phrase. The subject must then be an antitopic, and it will receive a characteristic flat intonation. That is, the two varieties in (8) are exactly parallel to the two options that were noted with respect to discourse functions in (2). I therefore conclude that the syntactic structure of (8a) is like the one shown in (5), with FinP moved to Spec-ForceP via Spec-TopP, and with the clause-final subject sitting in Spec-FocP, whereas the (8b) has the syntactic structure that was shown in (6), with the subject in the Spec of the lower TopP and the FinP raised to Spec-FocP and Spec-ForceP.

2.2 Clause-final subjects and reconstruction

A possible objection to my analysis of (8a) is that the focused constituent is not within the scope of the negation if the surface structure is as shown in (5). The negation will be too deeply embedded in FinP to be able to c-command the constituent in Spec-FocP. However, it is easy to show that a focused element is allowed to move out of the scope of the negation in the syntax while still being interpreted within that scope. Consider (9), where the negation clearly targets the fronted focus phrase. This is not very surprising given that the movement that the focus phrase has undergone is Aʹ-movement, which generally allows reconstruction. Thus, at the semantic interface the focus phrase is within the scope of the negation.

1. (9)

The structure in (5) is admittedly more complicated. Not only has the focus expression been extracted out of FinP, but FinP has also moved across the extracted phrase. Hence, interpreting the focused element as being inside the scope of the negation would require not only the focused phrase but in fact the whole FinP to reconstruct, that is, to be interpreted in the position where it appeared before movement. I will argue that this is what happens.

First, note that in Norwegian a clause-final subject can bind into a preceding constituent, as in (10a), but as (10b) shows, the converse is not possible. The example in (10c) is to show that under normal circumstances, the possessor in the partitive possessive construction can indeed be reflexive, so the problem with (10b) must lie elsewhere.

1. (10)

The binding relation in (10a) and the unavailability of the reflexive in (10b) might be taken to indicate that the clause-final subjects have right-adjoined in a position which is high enough to allow it to bind into the FinP. However, certain facts from Turkish suggest that this is not the case.

Like many other verb-final languages, Turkish allows various constituents, even subjects, to appear in postverbal position. An example is given in (11) (from Kornfilt Reference Kornfilt1997:206).

1. (11)

As indicated, the postverbal subject has a marked status in the discourse. Kornfilt (Reference Kornfilt1997:206) says that postverbal constituents in Turkish represent the shared presuppositions of the speaker and the hearer, while Erguvanlí (Reference Erguvanlí1984:50ff.) and Kural (Reference Kural1997:499) characterise their discourse function as background. Either formulation seems to be compatible with the postverbal constituents being antitopics, which is also a possible reading of Scandinavian clause-final subjects, as we have seen.

Since the unmarked order in Turkish is SOV, it is more evident in Turkish than in Norwegian that a clause-final argument must have left the position where it was first merged. Now as Kural (Reference Kural1997) notes, it is possible for a postverbal anaphor or pronoun in Turkish to be bound from a preverbal position. This is shown in (12) (from Kural Reference Kural1997:506).

1. (12)

According to Kural (Reference Kural1997), Turkish postverbal constituents are right-adjoined to CP, and they undergo reconstruction at LF whenever it is required by the binding relations, but not otherwise. We would then expect there to be no reconstruction if appropriate binding relations can be established without reconstruction. This means that if Kural's proposal is correct, a postverbal object should be able to bind into a preverbal subject. But as (13) demonstrates, this is not possible (Murat Kural, p.c.).

1. (13)

Whereas the ungrammaticality of (13) requires some additional stipulation if Kural's (Reference Kural1997) analysis is adopted, it follows naturally from the analysis I am proposing. After the object and the FinP have moved separately to the C-domain, there is no c-command relation between the subject and the object. A binding relation between these arguments must therefore be established before movement to the C-domain. But at that stage, the object is lower than the subject, and consequently, binding into the subject by the object is impossible whether or not there is reconstruction.Footnote ³

In (12a, b), on the other hand, the binding relations that we see are compatible with the structural relations before movement to the C-domain. Hence, if the word order at Spellout is derived as indicated in (5), it must be the case that at the semantic interface, both the fronted FinP and the postverbal constituent are read off in the positions where they appeared before they moved to the C-domain. That is, at the semantic interface the FinP is back in its position below the C-domain, and the object is back inside FinP. Now if this is possible in Turkish, it should be possible in Norwegian as well, and accordingly, I maintain that (8a) is derived as shown in (5), and that the binding relations in (10a, b) involve reconstruction.

2.3 Other properties of Scandinavian clause-final subjects

Let us now consider the Norwegian paradigm given in (14), which is representative of Scandinavian more generally. We see that an indefinite plural subject positioned in Spec-FinP, as in (14a), can have a generic reading (Gen) or, perhaps with some effort, an existential (Ex) reading. But with an expletive in Spec-FinP, its associate, the lexical subject, can only have an existential reading, regardless of whether it appears in its FinP-internal position, as in (14b), in a focus position at the left periphery, as in (14c), or in clause-final position, as in (14d).

1. (14)

The readings in (14) can be explained as follows. Generic indefinite subjects in Scandinavian must move to Spec-FinP, as in (14a). Existential indefinite subjects can also move to this position, or alternatively appear in a lower (possibly vP-internal) position. Both findings are consistent with the generalisations in Diesing (Reference Diesing1992). When the lexical subject appears in the lower position, Spec-FinP is obligatorily filled by an expletive, as in (14b). When a subject in the lower position is fronted across the expletive, as in (14c), the existential reading is retained. This is also the case if the FinP is fronted across a subject that has been so extracted, as in (14d).

In Scandinavian, the FinP can move across an extracted subject only if there is an expletive in Spec-FinP. But the expletive appears in Spec-FinP only if the lexical subject itself has not moved to Spec-FinP. Hence, a clause-final subject must have been extracted from a position below Spec-FinP. This is reflected in the fact that unlike a subject in Spec-FinP, but like a subject in the lower FinP-internal position, a clause-final subject must be indefinite – compare the examples in (14) to (15a, b, d). The same holds for a subject that has been fronted across an expletive – compare (14c) and (15c).

1. (15)

In other words, it seems that subjects can be definite only if they move to Spec-FinP at some point of the derivation. But note that it is arguably not the case that there must be an expletive in Spec-FinP whenever the lexical subject is not in that position. In (16) the clause-initial subject has probably raised to Spec-ForceP via Spec-FinP (since there is no expletive in that position) and Spec-FocP (since the subject is focused). And just like in (14a), it can have a generic or an existential reading. That is, either reading is allowed as long as the subject has touched down in Spec-FinP, which is where the generic reading arises.

1. (16)

If the subject in (16) is in Spec-ForceP, there must be a trace or, in other words, a copy that is not spelled out, in Spec-FinP. This is expected to be possible as long as Spec-FinP is c-commanded by the moved subject. The difference between (16) and the example in (14d), where the subject is also in the C-domain, according to my analysis, is that in the latter case, the FinP is shifted leftwards over the subject. The result is that the subject does not c−command Spec-FinP, and consequently, the expletive becomes obligatory.

To be more explicit, raising of FinP to Spec-XP is allowed in the configuration in (17a), but not in (17b).

1. (17)

   1. a. [_XP X [_YP Subj Y [_FinP EXPL . . . Subj . . .]]]

   2. b. [_XP X [_YP Subj Y [_FinPSubj . . .]]]

In (17a), the FinP-internal subject copy is licensed through c-command by the expletive in Spec-FinP, and it will still be c-commanded by the expletive even if FinP moves to a higher position. Hence, the structure is still legitimate after raising of FinP across the extracted lexical subject. In (17b), on the other hand, the lower subject copy is c-commanded by the higher subject copy. In this configuratioin, raising of FinP will result in an unlicensed copy in Spec-FinP, since this copy will then end up in sentence-initial position and it will not be c-commanded by another constituent in subject position.

2.4 On expletives in Scandinavian

It should be noted that the expletive det, seen in several of the preceding Norwegian examples, is formally identical to the third person singular neuter personal pronoun. This can be seen from (18).

1. (18)

Det is also the subject of weather verbs (i.e. it functions as a quasi-argument), and it fills the canonical subject position when the lexical subject is a postverbal clause. This is shown in (19).

1. (19)

These facts indicate that det has a full set of nominal features, and it can value the features of Fin without help from a nominal associate.

Swedish det has the same properties as Norwegian det – in fact, examples (18) and (19a) could also be Swedish, and in (19b), only a few words would be spelled differently. In Faroese, the form of the pronoun/expletive is tað, and in Icelandic, it is það. In (20), I show that the Icelandic það is the subject of weather verbs, it fills the subject position when the lexical subject is a postverbal clause, and it is also used as an expletive in presentational constructions with clause-final subjects.

1. (20)

With weather verbs, Danish also uses the pronominal quasi-argument det, as the example in (21a) illustrates. But as (21b) shows, as an expletive with a clausal subject associate, det alternates freely with der (see Vikner Reference Vikner1995:243–244). Der parallels the English there in that it is otherwise a locative adverbial.

1. (21)

When the associate is a DP, on the other hand, only der is possible:

1. (22)

Nevertheless, expletive der-constructions in Danish have the same syntactic properties as other Scandinavian expletive constructions. While a non-expletive DP in the canonical subject position can have a generic or an existential interpretation, the DP associate of der only has an existential interpretation, regardless of its linear position relative to the expletive. Moreover, the associate of der must be indefinite. These properties are illustrated in (23) and (24).

1. (23)

2. (24)

These data indicate that the subject in presentational expletive constructions in Danish, when it appears outside FinP, has not moved via Spec-FinP. Instead, it has crossed the expletive that sits in Spec-FinP. Moreover, the Danish expletive der has the same syntactic properties as the Norwegian/Swedish expletive det.Footnote ⁴

3.1 Some properties of clause-final subjects in English

In English, clause-final subjects are normally associated with focus – see, for example, Rochemont & Culicover (Reference Rochemont and Culicover1990:25). Thus, there is a clear contrast between the sentences in (25a) and (25b).

1. (25)

   1. a. There came to my office a horde of angry students after the meeting.

   2. b. There came to my office after the meeting a horde of angry students.

In (25a), no particular presuppositions need be invoked, and the focus domain may extend over the whole proposition. In (25b), the displaced subject phrase stands out as where the assertion differs from the presupposition. That is, the subject is unambiguously focused. This indicates that the structure in (5) is available in English as well.

A more remarkable property of English clause-final subject constructions is that there is no definiteness effect. In (26), the definite subject is acceptable (see e.g. Rochemont & Culicover Reference Rochemont and Culicover1990:29). By contrast, when the subject is in a low FinP-internal position, as in (27), it must be indefinite.

1. (26) There walked into the room the man she had no desire to see.

2. (27)

   1. a. There was a man walking into the room.

   2. b. *There was the man walking into the room.

The facts are similar when it comes to expletive passive constructions. If the stative reading is ignored, the real argument in an expletive passive construction cannot be definite if it is spelled out below Spec-FinP, as in (28a). But if it is in Spec-FinP, as in (28b), or in clause-final position, as in (28c), it can be definite.

1. (28)

   1. a. *There was the book placed on the table.

   2. b. The book that Mary wanted everyone to see was placed on the table.

   3. c. There was placed on the table the book that Mary wanted everyone to see.

The absence of definiteness effect in (26) and (28c) is clearly a problem for the analysis given in Chomsky (Reference Chomsky and Kenstowicz2001), according to which a phonological operation puts the lexical subject in clause-final position in constructions of this type. This means that (26), for example, is derived from (29) by an essentially phonological operation.

1. (29) *There walked the man she had no desire to see into the room.

As we see, (29) is completely ungrammatical, and the reason is that the lexical subject is definite and appears in a position where definite DPs are not allowed. If the definiteness effect has to do with licensing conditions for DPs (see e.g. Belletti Reference Belletti1988), it is rather unlikely that a non-syntactic operation can make the subject escape the requirement that it should be indefinite. Instead, the operation must be syntactic.

3.2 The relation between the expletive and its associate

Another fact about English expletive constructions is that the associate may not c-command the expletive. Consider the English paradigm in (30), which parallels the Scandinavian paradigms given in (14) and (23).

1. (30)

We see that English is like Scandinavian in that a full subject in the canonical subject position can have a generic or an existential interpretation, while only the existential interpretation is available for the associate of an expletive. The difference between English and Scandinavian is that the English (30c), with a topicalised associate c-commanding the expletive, is ungrammatical, whereas the corresponding Scandinavian constructions in (14c) and (23c) are fine. But the ungrammaticality of (30c) cannot be blamed on the properties of there as such, since, as we have seen, Danish uses a similar expletive in exactly parallel constructions. Instead, it must be the relation between the expletive and its associate that is different in English.

Let us nevertheless first look at the properties of the expletive. Just like the Danish expletive der, English there is always coindexed with an associate DP. When there is no associate DP, there cannot fill the subject position in English. Thus, we find it and not there as the subject of weather verbs, as in (31a), and where the lexical subject is a clause, as in (31b).

1. (31)

   1. a. {It/*there} is raining.

   2. b. {It/*there} was said that he was guilty.

The impossibility of having expletive there in (31a, b) follows if there does not have a full set of nominal features. I will assume, with Chomsky (Reference Chomsky and Kenstowicz2001:16), that there has only a person feature. The number feature must then be filled in by a fully specified DP. A well-known consequence is that the verb shows number agreement with the associate in constructions with expletive there.

So far, English there is like Danish der. But as we have seen, English expletive constructions are different from their Danish counterparts.Footnote ⁵ My proposal is that the differences are due to a special property of English expletive constructions, namely, that while Scandinavian expletives are merged in positions that the associate has not moved to, the English expletive there can be inserted in positions that the associate has moved out of. In other words, there can spell out the person feature of a copy of the associate.Footnote ⁶ However, for economy reasons, this may only happen when it is necessary for convergence. The properties of English expletive constructions now follow.

Firstly, if there can spell out a copy of its associate, the absence of definiteness effect in constructions like (26) is expected. On my analysis, (26), repeated below as (32a), is derived by moving the lexical subject to Spec-FinP, which is the only FinP-internal position where definite subjects can appear. After that, the definite subject moves out of FinP to Spec-FocP, and then the FinP raises to a topic position above where the subject has ended up. The result is the syntactic structure shown in (32b).Footnote ⁷

1. (32)

   1. a. There walked into the room the man she had no desire to see.

   2. b.

In this structure, the subject does not c-command its copy in Spec-FinP. Hence, English does not allow that copy to be phonologically null, and consequently, it is spelled out as there instead.

Secondly, if there has only a person feature it must be a third person feature. It follows that the associate of there cannot be first or second person. Hence, the construction in (33), with a first person pronoun in the clause-final subject, is much worse that (32a).

1. (33) *There walked into the room me/I and my two faithful followers.

Now consider the example in (34a), which shares with (32a) the movement of the lexical subject to Spec-FinP and then to a focus position in the C-domain, but differs in that the subsequent raising of FinP has not taken place.

1. (34)

   1. a. Poorly motivated students (*there) rarely come to my lectures.

   2. b. [_FocP Subj Foc [_FinPSubj . . . Subj . . .]]

   3. c. *[_FocP Subj Foc [_FinPthere . . . Subj . . .]]

As indicated, it is not grammatical to spell out the subject copy in Spec-FinP as there in this case. Since the higher subject copy c-commands Spec-FinP, it is OK to leave Spec-FinP phonologically empty, as in (34b), and consequently, spelling it out as there, as in (34c), is not allowed.

3.3 English clause-final subjects and genericity

Another contrast between constructions like (34a), with the focused subject in clause-initial position, and constructions like (26) and (30d), with the focused subject in clause-final position, is that the subject can have an existential or a generic reading in the former but only an existential reading in the latter. I sum this up in (35).

1. (35)

Intuitively, the contrast seen here is unexpected on my analysis. If the subject has moved via Spec-FinP, we would expect it to retain a possible generic reading whenever it raises to Spec-FocP, and we would also expect the raising of FinP across the extracted subject to have no influence on the interpretation of the subject. So why does (35b) not allow a generic reading of the subject?

It has been observed before that the presentational there-construction, exemplified by (35b), is only possible with stage level predicates. This is noted by Chierchia (Reference Chierchia, Carlson and Pelletier1995), who proposes that individual level predicates involve a generic operator which competes with the existential operator supplied by the there-construction in such a way that the two are incompatible. He also points out that individual level predicates, but not stage level predicates, assign a generic reading to bare plural subjects. It follows that bare plural subjects in there-constructions cannot have a generic reading.

Inspired by Milsark (Reference Milsark1979), Chierchia suggests that the generic operator supplied by individual level predicates turns the subject into a strong NP. Hence, in expressions like (36) we have something very similar to the definiteness effect: since it is bound by the generic operator of the predicate tall, the subject a man is a strong NP, and consequently, it is not acceptable in there-constructions.

1. (36) ??There is a man tall.

Note, however, that (36) is an example of an existential expletive construction, whereas (35b) is a presentational expletive construction. It has been known since Aissen (Reference Aissen1975) that the two construction types are different. For example, as we have seen, the associate in presentational expletive construction can be definite. This means that it is not the resulting strength that prevents the subject in (35b) from receiving an individual level interpretation. Nevertheless, the generalisation that the presentational expletive construction only allows stage level predicates seems to hold, whatever the reason might be. Since a generic reading of the subject in (35b) would require an individual level predicate, a generic reading is not available here.

It is not the case, though, that the presentational there-construction disallows genericity in general. If we take (35b) as an example, we see that this sentence can be interpreted generically with a reading that can be rendered informally as ‘for my lectures it holds that poorly motivated students rarely come to them’, although the reading ‘for poorly motivated students it holds that they rarely come to my lectures’ is not available. In terms of Krifka et al. (Reference Krifka, Pelletier, Carlson, ter Meulen, Chierchia, Link, Carlson and Pelletier1995:25), the sentence allows a reading where my lectures is the restrictor, but not a reading where poorly motivated students is the restrictor. The deeper reason for this contrast will have to be worked out.

3.4 Presentational there and locative inversion

Given the analysis I have proposed of presentational constructions with clause-final subjects, a question arises concerning constructions with locative inversion, as exemplified in (37).

1. (37) Into the room walked John.

Locative inversion is known to have several properties in common with presentational there-constructions. For example, Rochemont & Culicover (Reference Rochemont and Culicover1990) point out that the subject is postverbal and focused in both constructions, and moreover, neither construction allows wh-movement of the subject or out of the subject. Furthermore, locative inversion allows the subject to be definite, as (37) demonstrates. The same holds of a clause-final subject in the presentational there-construction, as we have seen. The similarities between presentational there and locative inversion might be taken to suggest that the two construction types are derived in a parallel fashion.

Now if my analysis of presentational there-constructions were to be extended to locative inversion constructions, it would mean that locative inversion is the result of raising the subject out of FinP and then raising the FinP, with an initial PP, to a position above the landing site of the subject. This is not totally implausible, since it has been proposed, by Hoekstra & Mulder (Reference Hoekstra and Mulder1990), Bresnan (Reference Bresnan1994), Collins (Reference Collins1997), and others, that the locative PP fills the canonical subject position in locative inversion constructions (but see Culicover & Levine Reference Culicover and Levine2001 for some refinement).

However, in spite of the observed similarities, there are also differences between the presentational there-construction and the locative inversion construction. Firstly, unlike clause-final subjects in presentational there-constructions, the subject in locative inversion constructions shows no heaviness effect. This is seen already in (37). Secondly, while the displaced subject in the presentational there-construction is strictly clause-final, as shown in (38), the subject in the locative inversion construction is not necessarily clause-final. It can, for example, be followed by a secondary predicate, as in (39) (the examples in (38) and (39) are from Rochemont & Culicover Reference Rochemont and Culicover1990:76–77).

1. (38)

   1. a. There walked into the room nude a man no one knew.

   2. b. *There walked into the room a man no one knew nude.

2. (39)

   1. a. *Into the room walked nude John.

   2. b. Into the room walked John nude.

These differences indicate that the two constructions are rather different after all. In particular, it appears that the clause-final constituent in locative inversion is not just the subject; it is a larger phrase containing the subject.

Given the facts listed above, I conclude that locative inversion differs from the presentational there-construction to such a degree that it falls outside the scope of this article.

4.1 Clause-final subjects and wh-movement in a single clause

An interesting fact pointed out by Chomsky (Reference Chomsky and Kenstowicz2001) is that in a clause with a subject in final position, wh-movement is possible as long as it does not apply to the subject. This is shown for English in (41) and for Norwegian in (42). In each example, there is a wh-moved locative phrase (in boldface) in clause-initial position and a subject (in small capitals) in clause-final position.

1. (41) At which airport did there arrive yesterday three strange men?

2. (42)

If the postverbal subjects in these examples are in a Spec position in the C-domain, as I am claiming, more precisely in Spec-FocP, they must nevertheless be in a position different from where the wh-phrases are located. This is unexpected on the common assumption that wh-movement and focus movement target the same position. However, Tsimpli (Reference Tsimpli, Joseph, Horrocks and Philippaki-Warburton1998) shows that in Greek, focus movement and wh-movement may co-occur in one single embedded clause. The following example, with a focused phrase followed by a wh-moved phrase inside an indirect question, is from Tsimpli (Reference Tsimpli, Joseph, Horrocks and Philippaki-Warburton1998:204).Footnote ⁸

1. (43)

In Greek, a focused phrase and a wh-phrase may not co-occur at the left periphery of a matrix clause, though. The explanation that Tsimpli (Reference Tsimpli, Joseph, Horrocks and Philippaki-Warburton1998) suggests is that matrix clauses are smaller than embedded clauses, so that in matrix clauses, focused phrases and wh-phrases compete for one single position. But there appears to be another possible explanation for the observed fact. As Tsimpli notes, when focus fronting and wh-movement co-occur, the focused phrase must precede and take scope over the wh-phrase. On the other hand, she assumes that interrogative type-marking of a matrix clause depends on the movement of the wh-operator. An embedded interrogative clause, by contrast, is type-marked by selection. This must mean that having a focused phrase in front of a wh-phrase does not interfere with the type-marking of an embedded clause. For matrix clauses, however, it is likely that the wh-operator must be the highest operator if the clause is to count as interrogative. Hence, a fronted wh-phrase cannot be preceded by a focused phrase in a matrix clause. But since a phrase that has undergone focus movement cannot follow the wh-phrase either, it follows that focus movement and wh-movement cannot both apply in one single matrix clause.Footnote ⁹

Nevertheless, I will hypothesise that the positions for focused phrases and for wh-phrases that we see in (43) are in principle available in all clauses. If so, it should be possible to derive (44b) from (44a) by moving three strange men to a focus position above the position where the fronted wh-phrase is located. However, (44b) is not well formed, since the wh-operator is not the highest operator. But since the wh-operator cannot take scope over the focus operator, it is not possible to move the wh-phrase across the focused constituent, for example to Spec-ForceP. The only way to save (44b) is by moving the constituent containing the wh-phrase and the remainder of FinP – that is, QP – to the Spec of ForceP. This is sketched in (44c).

1. (44)

In this configuration, the wh-operator is higher than the focus operator, but the wh-phrase does not c−command the focused phrase. Thus, the structure in (44c) is licit. Moreover, the constituent in Spec-FinP is now spelled out as there, since it is no longer c-commanded by the raised associate. My proposal is therefore that (41) is derived as indicated in (44), and that (42) is derived in a parallel fashion, except that the expletive sits in Spec-FinP from the beginning and the lexical subject moves across it.

A fact that needs some comment is that both (41) and (42) are V2 clauses. In these cases it is likely that V2 is induced by Q, the head that attracts the wh-phrase to its Spec. Q also attracts the finite verb, so that QP becomes headed by the finite verb. Consequently, it is the whole QP that is attracted to ForceP by the finiteness feature of Force. Since QP is a phrase and not a head, it ends up as the specifier of ForceP, thereby also satisfying the EPP feature of ForceP.

For some speakers of English, it is also possible to have polarity questions with clause-final subjects, as in (45).Footnote ¹⁰

1. (45) %Did there run into the room several overexcited fans?

We now see that they can be derived as sketched in (46), on the assumption that the fronted auxiliary in English polarity questions raises to the head of QP.

1. (46)

After the FinP is completed, as in (46a), Q is merged, and the finite verb did is attracted to Q – see (46b). Then a Foc head is merged over QP, and the focused subject, several overexcited fans, moves to Spec-FocP, as shown in (46c). Finally, Force is merged, and QP, headed by the finite verb, moves to Spec-ForceP. In the resulting configuration, the (person feature of the) subject copy in Spec-FinP is spelled out as there – see (46d).

4.2 Wh-movement applying to the clause-final subject in English

While wh-movement and focus movement are combined in the examples we have just considered, ungrammaticality results, in English, if we try to wh-move a clause-final subject or an element belonging to it. The impossibility of wh-moving the clause-final subject as a whole was demonstrated in (40), which I repeat here as (47).

1. (47)

   1. a. There arrived in the mail five heavy packages.

   2. b. *How many packages did there arrive in the mail?

Wh-movement out of a clause-final subject is shown in (48). We see that (48a), with a clause-final subject, is relatively acceptable, but wh-moving a phrase out of this subject is not possible – see (48b).

1. (48)

   1. a. ?There was delivered to my office a picture of Chomsky.

   2. b. *Who was there delivered to your office a picture of who?

Recall that Chomsky (Reference Chomsky and Kenstowicz2001) takes the ungrammaticality of the constructions in (47b) and (48b) as evidence that movement of the subject to clause-final position is a phonological operation. If it is a phonological operation, it follows that wh-movement cannot subsequently apply to the subject. But notably, it is not necessarily the case that the wh-phrase in (47b) has been extracted from clause-final position. It could have been extracted from the position where it was originally merged. In other words, the position of the wh-copy could be as in (49b) rather than (49a).

1. (49)

   1. a. *How many packages did there arrive in the mail how many packages?

   2. b. *How many packages did there arrive how many packages in the mail?

In Chomsky (Reference Chomsky and Kenstowicz2001) it is also suggested that there is a requirement in English that the single argument of a passive or unaccusative verb should move out of VP, and that this is why that argument tends to appear in clause-final position when the canonical surface subject position is filled by an expletive. One might then wonder why wh-movement alone does not suffice to meet the requirement that the argument should move out of VP. If this were the case, (47b) should be grammatical, contrary to fact. Hence, we have identified a problem with Chomsky's analysis.

The right solution is, in my view, that the ungrammaticality of the constructions in (47b), (48b) and (49) has nothing to do with the subject having moved to clause-final position before it wh-moves. It is a consequence of the presence of the expletive there in a position that is c-commanded by the wh-moved associate. Since Spec-FinP is c-commanded by and coindexed with the wh-phrase, it is licit to have a phonologically empty constituent in Spec-FinP. Accordingly, Spec-FinP cannot be realised as there. In other words, the impossibility of applying wh-movement to the associate in English presentational expletive constructions has the same source as the ungrammaticality of (30c), where the expletive in Spec-FinP is c-commanded by the associate in Spec-FocP.

Now consider how the derivation of (48b) would proceed, if the suggestions in (44) are correct. The derivation is shown in (50). In (50a), a picture of who has raised to Spec-FinP. Then who is extracted by wh-movement and the finite verb moves to Q – see (50b). The DP containing the copy of who is then raised across the wh-phrase to Spec-FocP – see (50c). The final step is raising of QP, containing the wh-phrase and the rest of FinP, to Spec-ForceP, across the focused phrase – see (50d).

1. (50)

In this case, the whole associate of the expletive is present in Spec-FinP before Spellout. That is, the wh-phrase is underlyingly present not only in the position from where it appears to be extracted, but also in the constituent in Spec-FinP. Hence, if the constituent in Spec-FinP is spelled out as there, there will be c-commanded by a constituent (the fronted wh-phrase) which is also included in the constituent that is spelled out as there. The ungrammaticality of the construction indicates that presentational there resists being c-commanded by any element from its underlying representation, that is, by its associate or by a part of its associate.Footnote ¹¹

The analysis I have suggested of expletive there can also explain the contrast shown in (51). In (51a) as well as in (51b) there is wh-extraction out of a complement in an indefinite DP. However, the active (51a) is grammatical but the passive (51b) is not.

1. (51)

   1. a. What are they selling books about what in Boston these days?

   2. b. *What are there books about what being sold in Boston these days?

On an analysis along the lines of Chomsky (Reference Chomsky and Kenstowicz2001), the reason for this must be that the DP that the wh-phrase is extracted from is in its base position in (51a) whereas in (51b), it has to get out of the VP, and this happens before wh-movement applies. This leaves open the question of why wh-movement from within the DP cannot apply before the DP itself is moved. This would be expected if the movement of the DP out of VP is a phonological operation, and it would derive a grammatical (51b). If, on the other hand, the expletive there in presentational constructions is the partial spellout of a DP-copy, as I am proposing, then the syntactic structure of (51b) is as sketched in (52).

1. (52)

Here again, there is c-commanded by an element, the highest what, that also is part of the phrase that there spells out a copy of. This is why the construction is ungrammatical.

But note that existential expletive constructions in English do allow wh-movement to apply so that there ends up being c-commanded by its associate, as in (53).Footnote ¹²

1. (53) What is there left in the fridge?

The difference between existential expletive constructions and presentational expletive construction with respect to wh-movement was first pointed out by Aissen (Reference Aissen1975), who also listed a number of additional differences between the two constructions. In our present context, the more interesting of these differences is that an existential expletive construction can be embedded under any predicate that takes a propositional complement, whereas presentational expletive constructions can only be embedded under predicates that allow root phenomena in their complements. Thus, a factive verb like regret can embed an existential expletive construction but not a presentational expletive construction. This is illustrated in (54a, b) (Aissen's examples (24a) and (22b)).

1. (54)

   1. a. We regret that there is no possibility of a job here.

   2. b. *The driver regrets that there stepped out in front of his car a pedestrian.

This suggests that the presentational expletive construction involves the higher part of the clause whereas the existential expletive construction does not. It is, then, not likely that the operation that puts the subject in clause-final position in the presentational expletive construction is phonological or only affects the lower part of the clause. The facts are better explained by my analysis, which takes the word order of the presentational expletive construction to involve high heads that may not even be present in embedded clauses without root properties.

Aissen's account of the difference between the existential construction and the presentational construction is that the latter results from a root transformation, and because of this the whole clause is in principle an island. Existential constructions, by contrast, do not involve root transformations and are not islands. Now for my explanation of the ungrammaticality of wh-movement applying to the associate in presentational expletive constructions to go through, it is necessary that the expletive in existential expletive constructions is not the partial spellout of a copy of the associate. If it were, (54a) would also be predicted to be ungrammatical. Several analyses have been proposed that could be relevant here, for example the analysis in Williams (Reference Williams1994:135), according to which the associate in existential expletive constructions is a predicate, and the expletive is structurally its subject. The analysis in Hazout (Reference Hazout2004) is nearly identical (as Williams Reference Williams2006 points out). But if the expletive is the structural subject of its associate, it is not surprising that the associate can be wh-moved across the expletive to a position where it c-commands it. This would be parallel to wh-moving any predicate across its subject. Hence, if Williams (Reference Williams1994) is right, the grammaticality of (54a) is no problem for my analysis of wh-movement in presentational expletive constructions.

4.3 Clause-final subjects and wh-movement in Scandinavian

Potential support for my analysis of the English constructions that we have just looked at is found in Scandinavian. In Norwegian, for example, the counterpart to (47b) is perfectly grammatical, as shown in (55).

1. (55)

Norwegian also has grammatical constructions that are not very different from (51b):

1. (56)

One difference between the Norwegian (56) and the English (51b) is that wh-extraction in (56) no doubt takes place from a phrase that has not moved, whereas in (51b), the lexical argument moves to a position in front of the auxiliary being before extraction takes place. One might want to claim that (A-)movement blocks subsequent wh-movement, and that this is the main reason for the contrast in grammaticality between the English examples in (47b) and (51b) on the one hand and the Norwegian examples in (55) and (56) on the other. In the Norwegian examples, wh-movement targets a constituent that has not moved previously, but in the English examples, the relevant constituent undergoes movement before wh-movement (unsuccessfully) applies. But, as we will see in a moment, the relevant contrast has to do instead with the expletive.

But let us first note that certain facts about Scandinavian might also lead to the conclusion that wh-movement in expletive constructions cannot apply to associates that have moved. In Swedish and in some varieties of Norwegian, both the order participle > object and the order object > participle is possible in expletive passive constructions (other varieties of Norwegian allow only the participle > object order), and while the participle then agrees with the object in gender and number in the object > participle order, the participle has the default neuter singular form in the participle > object order. The examples in (57) are Norwegian.

1. (57)

When the associate undergoes wh-movement, the participle does not normally agree with it, as illustrated in (58) (see Holmberg Reference Holmberg2001:108).

1. (58)

This fact suggests that wh-movement of the associate is only possible from the postparticipial position, not from the preparticipial position.

However, it appears that wh-movement of a part of a clause-final subject in a presentational expletive constructions is not totally out in Scandinavian. This is shown in (59) and (60). Both in (59), which is Norwegian, and in (60), which is Swedish, the lexical argument is clearly displaced, since it follows an adverbial that follows the nonfinite verb. The declarative versions (59a) and (60a) are fully grammatical. The corresponding questions in (59b) and (60b) receive judgements that range from ‘perfectly OK’ to ‘very clumsy’, but they are never deemed totally ungrammatical. (As indicated, my Swedish informants are slightly more positive than my Norwegian informants, but this might be purely accidental, due to the small number of informants.)

1. (59)

2. (60)

The derivation of (60b) is shown in (61), starting from the point where FinP is completed and has the expletive det in its highest Spec, as shown in (61a). After that, Q is merged, the wh-word vad ‘what’ moves from within the nominal argument to Spec-QP, and the finite verb moves to Q – see (61b). Then Foc is merged, and the remainder of the nominal argument moves to Spec-FocP – see (61c). Finally, Force is merged, and QP, now headed by the finite verb, moves to Spec-ForceP – see (61d).

1. (61)

The resulting construction is a little hard to parse, harder than its declarative counterpart, since there is a wh-dependency as well as a focus dependency, and the question word vad is involved in both dependencies. Similar problems are felt in (59b). But despite the parsing difficulties the construction is in principle allowed in Scandinavian. The fact that (59b) and (60b) are structurally parallel to the ungrammatical English example in (48b) then tells us that the reason why (48b) is ungrammatical is not that illicit movements have taken place, as Chomsky (Reference Chomsky and Kenstowicz2001) would have it. Rather, it is likely that the crucial difference between (48b) on the one hand and (59b) and (60b) on the other lies in the expletive. On my analysis, the English (48b) is ungrammatical because there ends up being c-commanded with an element that is a part of the underlying representation of there. But in Scandinavian, expletives do not spell out copies of moved constituents. This holds for der as well as for det. Hence, the fact that the fronted wh-constituent c-commands the expletive in the examples above does not lead to ungrammaticality. More generally, the problems that we see in English when an associate is wh-moved across an expletive do not arise in Scandinavian.Footnote ¹³

5.1 The divalent expletive construction in English

The English ‘transitive expletive’ construction, as Chomsky (Reference Chomsky and Kenstowicz2001) terms it, is exemplified in (62). As we see, the verb has two arguments, and in addition an expletive appears in the canonical surface subject position. The argument that seems to be the subject must follow what appears to be the object. Because of this, Chomsky (Reference Chomsky and Kenstowicz2001) concludes that rightward displacement of the subject is obligatory even here.

1. (62)

   1. a. There hit the stands a new journal.

   2. b. *There hit a new journal the stands.

   3. c. There entered the room a strange man.

   4. d. *There entered a strange man the room.

However, the ordering restrictions illustrated in (62) should not necessarily be taken to mean that there is a subject which is obligatorily displaced to the right. Rather, I would claim that the verbs that appear in the English ‘transitive expletive’ construction are not really transitive; that is, they have no external argument. As demonstrated in (63), verbs with external arguments cannot appear in the ‘transitive expletive’ frame.

1. (63)

   1. a. *There hit the stands a juvenile delinquent.

   2. b. *There boiled the eggs a new chef.

   3. c. *There boiled a new chef some eggs.

What we see in the grammatical (62a) and (62c) is a verb with two internal arguments possibly appearing in their base-generated order; that is, with either the goal argument or indirect object preceding the theme argument or direct object. This means that this is not really a ‘transitive expletive’ construction although it is an expletive construction with a divalent verb.

Significantly, it is possible, at least for some speakers, to have an adverbial following the two arguments in the divalent expletive construction, as in (64a). In this case, the direct object must be indefinite, hence the ungrammaticality of (64b).

1. (64)

   1. a. There entered the classroom a strange man yesterday.

   2. b. *There entered the room the Russian student the next minute.

Both the fact that the second argument in the English ‘transitive expletive’ construction can be followed by an adverbial and the fact that the second argument must then be indefinite suggest that both arguments are licensed within vP, or at least lower than Spec-FinP. In other words, there is no indication that there is obligatory displacement of any argument in this construction, and the claim to this effect put forward by Chomsky (Reference Chomsky and Kenstowicz2001) does not hold.

However, displacing the associate of the expletive to clause-final position is an option in the divalent expletive construction, in the same way as it is an option in unaccusative expletive constructions and passive expletive constructions. Strikingly, the definiteness effect then disappears:

1. (65)

   1. a. ?There entered the classroom yesterday a very strange man.

   2. b. ?There entered the room the next minute the new student from Russia.

My proposal is therefore that the constructions in (65a) and (65b) are derived by raising of the underlying direct object to the canonical surface subject position Spec-FinP, followed by raising of that argument to Spec-FocP and subsequent topicalisation of FinP, with the direct object copy in Spec-FinP spelled out as there. That is, my analysis of clause-final surface subjects of verbs with two internal arguments is similar to my analysis of clause-final subjects of unaccusative and passive verbs.

It follows that the reason why wh-movement of the associate of the expletive there is ungrammatical in the divalent expletive construction, as demonstrated in (66), is again that if the associate raises above there, it ends up c-commanding there, which is not allowed.

1. (66)

   1. a. *How many men did there enter the room?

   2. b. *How many journals did there hit the stands?

Given the analysis of (65a) and (65b) that I have just sketched, it is perhaps a little surprising that the subject of a true transitive verb cannot appear in clause-final position. In this respect, there is a sharp contrast between verbs with two internal arguments, shown in (67a) and (67b), and true transitive verbs, having an external and an internal argument, shown in (67c) and (67d).

1. (67)

   1. a. There hit the stands (yesterday) the new journal that Peter is editing.

   2. b. There entered the room (yesterday) the man she had no desire to see.

   3. c. *There hit the stands (yesterday) the gang that is terrorising the neighbourhood.

   4. d. *There boiled some eggs (yesterday) the new chef from France.

On my analysis, in every clause in (67) the associate of there has raised from Spec-FinP to Spec-FocP at one stage of the derivation, since it ends up following the adverbial. The ungrammaticality of (67c) and (67d) must then be taken to mean that a FinP containing a true transitive verb cannot be topicalised across a focused subject, at least not in English. It is not immediately obvious why this is so – if the expletive there can be the spellout of a subject copy in Spec-FinP, we would expect all the examples in (67) to have the same status.

Let us consider the derivation of (67a), which is shown in (68) below. Here the two arguments both follow the verb when the vP is completed, as seen in (68a). When Fin is subsequently merged, one alternative is to insert an expletive in Spec-FinP, which, without the adverbial, would give There hit the stands the new journal, an order that is similar to (62a) and (62c). This means that as long as there are no adverbials, we cannot tell whether the last argument is still inside the vP. Another alternative is to raise the second argument to Spec-FinP, as in (68b). In the case at hand, this argument is focused and moves on to Spec-FocP after Foc is merged – see (68c). After that, Top is merged, and the remnant FinP moves to Spec-TopP, as shown in (68d). And as indicated, Force is merged on top of TopP, but since the syntax of ForceP is not under discussion here, I leave it out (but see note 7).

1. (68)

Let us now compare (68) to the derivation of (67d), which is shown in (69) (the cross indicates where the fatal move occurs).

1. (69)

Here one argument (the agent) precedes the verb and one argument follows it when the vP is completed – see (69a). The agent also occupies Spec-FinP when FinP is completed – see (69b). Since it is focused, it moves on to Spec-FocP after Foc is merged, as shown in (69c). So far, we have derived structures that are perfectly fine. It appears that the ungrammaticality of (68d) arises when FinP is moved across the focused subject, as in (69d). The explanation I will offer is based on the proposal in Chomsky (Reference Chomsky and Kenstowicz2001) that a transitive v defines a strong phase v*P, and on the proposal in Fox & Pesetsky (Reference Fox and Pesetsky2004) that linearisation takes place cyclically, at the completion of each spellout domain. If we assume that the transitive v*P constitutes a spellout domain, while the intransitive vP does not, it follows that once the transitive v*P is completed, the relative linear order of the elements inside that v*P will be fixed once and for all. It is still possible to move the subject to a higher position, as in (69b) and (69c), since this movement will not alter the relative order of the subject and other elements inside v*P. Similarly, it will be possible to move further a wh-element or focused element that has got to the escape hatch in the initial position of v*P at the point when the v*P is completed. It is however forbidden to raise the remaining FinP over the subject, since this would reverse the ordering relation between the subject and all other elements inside v*P.Footnote ¹⁴ But when the verb has no external argument, as in (68), the vP does not constitute a spellout domain. The smallest spellout domain is ForceP. Hence, reordering of the verb and the arguments is allowed even after the vP is built. For example, one can extract one argument and then raise the whole FinP over it. The relative order of the elements that originate within vP will be fixed later, when the whole ForceP is completed.

5.2 Divalent expletive constructions in Norwegian

A divalent expletive construction similar to the English one is also found in Norwegian. Note first that Norwegian, unlike Icelandic, does not allow true transitive expletive constructions, regardless of the order of the nominal arguments. This is illustrated in (70).

1. (70)

Nevertheless, certain verbs, such as møte ‘meet’ and vente ‘await’, can be followed by two arguments, as demonstrated in (71).

1. (71)

We also see that the order of the arguments is fixed in these cases. Again, both arguments are arguably internal, on this particular use of the two verbs, and they appear in the order goal/IO > theme/DO (see Lødrup Reference Lødrup1995, Vikner Reference Vikner1995:219). Note that since the argument sequence can be followed by adverbials, it must be the case that both arguments can be spelled out inside FinP.

Just like in English, the underlying direct object of a divalent expletive construction can be displaced to clause-final position, as in (72a). However, as shown in (72b), the clause-final argument does not escape the definiteness effect in Norwegian.

1. (72)

The ungrammaticality of (72b) is expected given that a displaced subject in Norwegian expletive constructions has been extracted from a position lower than Spec-FinP. That is, the subject cannot have touched down in Spec-FinP – it must have been A’-moved across that position. It follows that the subject must be indefinite, and it also follows that transitive verbs, that is, verbs with one external and one or two internal arguments, never appear in expletive constructions. In constructions with transitive verbs, one argument must be licensed in Spec-FinP, and consequently, the expletive det cannot appear. Alternatively, since the Scandinavian element det does not need an associate, but can fulfil the requirements of T on its own (see (19) above), we must conclude that it has a full set of nominal features and that it can serve as a goal for the probe T. It is therefore likely that the expletive det starts out in Spec-vP, and again it follows that active transitive verbs cannot appear in expletive constructions.

Just like the presentational expletive construction, the divalent expletive construction allows wh-movement of the associate of the expletive in Norwegian:

1. (73)

Once more, Norwegian differs from English, and again, I take the difference to be due to the relation between the expletive and the associate. As I have argued above, the expletive det is not the spellout of a copy of the associate, and accordingly, it may be c-commanded by the raised associate. In English, as we have seen, a similar situation leads to ungrammaticality.