Abstract

In Wood 2012, I argued that object extraposition of infinitive clauses in Icelandic reveals a problem for the movement theory of control (MTC). Object extraposition involves a pronoun that, when present, prevents any movement out of the extraposed clause, but allows the control dependency. Drummond and Hornstein (2014) claim that the facts discussed in Wood 2012 are compatible with the MTC. In this reply, I show that their response is based on a misunderstanding of how Icelandic object extraposition works and that the problem observed in Wood 2012 remains. I also present a novel argument against the MTC by showing that Drummond and Hornstein’s account of ‘promise’ - type verbs cannot be extended to Icelandic.

1 Object Extraposition and the Movement Theory of Control

Thráinsson 1979 contains a detailed study of Icelandic clausal extraposition from the subject and object positions. In the course of this study, Thráinsson (1979) points out that control infinitives can extrapose from subject and object positions as well. In sentences like (1a–b), an optional (case-marked) pronoun það ‘it’ may appear; the presence of the pronoun often—but not always, as we will see below—indicates that extraposition has taken place.1

(1)

  • a.

    Þeiri ákváðu (það) að PROi heimsækja Ólaf.

    they.NOM decided (it.ACC) to visit Olaf.ACC

    ‘They decided to visit Olaf.’

    (adapted from Thráinsson 1979:111, 114)

  • b.

    Þeiri frestuðu (því) að PROi hálshöggva fangana.

    they.NOM postponed (it.DAT) to execute the.prisoners.ACC

    ‘They postponed executing the prisoners.’

    (adapted from Thráinsson 1979:228–229)

I argued in Wood 2012 that the availability of this pronoun in obligatory control structures is problematic for the movement theory of control (MTC) (Hornstein 1999, Boeckx and Hornstein 2006, Boeckx, Hornstein, and Nunes 2010a,b). According to the MTC, the surface subject in (1), þeir ‘they’, moves from the embedded subject position into the matrix clause by A-movement. The landing site of the movement is the external argument position of the matrix clause, Spec,vP. The control reading is derived because the A-chain contains one DP in two θ-positions: in this case, the embedded external argument position and the matrix external argument position.

However, Thráinsson (1979) shows that Ā-extraction can occur out of a control clause only when that pronoun is not present, as illustrated with topicalization in (2a).

(2)

  • a.

    Ólafi ákváðu þeir (*það) að PRO heimsækja ti.

    Olaf.ACC decided they.NOM (*it.ACC) to visit

    ‘Olaf, they decided to visit.’

  • b.

    Fanganai frestuðu þeir (*því) að PRO hálshöggva ti.

    the.prisoners.ACC postponed they.NOM (*it.DAT) to execute

    ‘The prisoners, they decided to postpone executing.’

Moreover, there are no known cases of A-movement past such pronouns. For example, (3) shows that aspectual verbs can have raising-verb syntax, since they preserve the dative case assigned by the embedded verb leiðast ‘feel bored’. (4) shows that the extra pronoun is not possible with these verbs.

(3)

  • Mér {byrjaði / hætti } að leiðast.

  • me.DAT {began / stopped} to bore

  • ‘I began/stopped feeling bored.’

  • (Sigurðsson 1989:70)

(4)

  • Haraldur {byrjaði (*það) / hætti (*því) } að senda henni bréf.

  • Harold.NOM {began (*it.ACC) / stopped (*it.DAT)} to send her letters

  • ‘Harold began to send/stopped sending her letters.’

(5) shows that these verbs are capable of assigning case to DP objects: byrja ‘begin’ assigns accusative and hætta ‘stop’ assigns dative. Thus, the problem with the extra pronoun in (4) cannot be that the matrix verb has no way to assign case to it.2

(5)

  • Þjálfarinn {byrjaði leikinn / hætti leiknum }.

  • the.coach.NOM {began the.game.ACC / stopped the.game.DAT}

  • ‘The coach began/stopped the game.’

My claim in Wood 2012 is that since a control dependency is possible past this pronoun, but no movement dependencies are possible past it, the control dependency should not be modeled as movement as it is currently understood. Of course, the MTC position has repeatedly emphasized that differences between raising constructions and control constructions are not automatically problems for the MTC, since “raising A-movement” and “control A-movement” involve different landing sites; therefore, differences between control and raising constructions could stem from these differences in the landing sites. I considered and dismissed two possible analytical avenues for the MTC, one treating these constructions as nonobligatory control and the other claiming that the pronoun was not in a θ-position and therefore not an intervener. Since these points have been left intact by Drummond and Hornstein (2014) (henceforth D&H), I will not discuss them further here.

D&H claim that the facts in Wood 2012 can be handled by the MTC under the assumption that extraposition involves right-adjunction to vP. The idea is this. The problem with (4) is that the DP moves to Spec,TP, but the extraposed infinitive is adjoined to vP. There are two ways of accomplishing this, and both lead to problems. First, one might adjoin the infinitive to vP and, from there, raise the subject out of the infinitive into the matrix subject position (Spec,TP). But this is impossible because an extraposed infinitive is an island for extraction. Second, one might first raise the subject out of the infinitive into the matrix Spec,TP and then adjoin the infinitive to vP. This, however, violates the Extension Condition, which requires that Merge apply to the root of the tree (Chomsky 1995). Since vP is dominated by TP, the clause cannot adjoin to vP after TP has been built. Control clauses, however, are different in that the DP in the infinitive is moving to the matrix Spec,vP, not to Spec,TP. So a subject can move to the matrix Spec,vP, and the clause can then adjoin to vP. The problems associated with raising clauses vanish: movement occurs before the clause becomes an island, and adjunction respects the Extension Condition.

I will show that D&H’s reply is on the wrong track in its attempt to account for control past the pronoun in terms of the landing site of object extraposition. The presence of the pronoun does not force extraposition, and clauses occurring with the pronoun are islands for extraction whether extraposition takes place or not. Therefore, movement is impossible even before extraposition takes place, and the landing site is irrelevant. Moreover, extraposition does not always force the presence of the pronoun, and when the pronoun is not present, extraposed clauses are not necessarily islands for extraction. Thráinsson (1979) argues in detail that the pronoun underlyingly forms a constituent with the clause and that this constituent is an island. I will show that the facts motivating Thráinsson’s analysis carry the same force today that they did in the theory current in 1979.

2 Constituency

D&H note that Thráinsson (1979) analyzes sentences such as (1a–b), when the pronoun is present, as involving object extraposition, but they do not discuss what Thráinsson’s analysis of object extraposition is. Thráinsson argues at length that the pronoun and the following clause form a constituent and that when object extraposition takes place, the clause is extraposed out of this complex NP.3Thráinsson (1979:219–220) cites numerous constituency tests in favor of this conclusion. They include topicalization (6a), left-dislocation (6b), ‘it’-clefting (6c), passivization (6d), and right-node raising (6e).

(6)

  • a.

    [Það að María skuli hafa farið] harma ég ákaflega.

    [it that Mary should have gone] regret I deeply

    ‘That Mary went, I deeply regret.’

  • b.

    [Það að María skuli hafa farið], það er hörmulegt.

    [it that Mary should have gone] that is deplorable

    ‘That Mary went, that is deplorable.’

  • c.

    Það er [það að María skuli hafa farið] sem Jón harmar ákaflega.

    it is [it that Mary should have gone] that John regrets deeply

    ‘It is that Mary went that John deeply regrets.’

  • d.

    [Það að María skuli hafa farið] er almennt harmað.

    [it that Mary should have gone] is generally regretted

    ‘That Mary went is generally regretted.’

  • e.

    Jón harmar en Haraldur elskar [það að María skuli hafa farið].

    John regrets but Harold loves [it that Mary should have gone]

    ‘John regrets but Harold loves that Mary went.’

I will refer to clauses headed by this pronoun as pronoun-headed clauses (PHCs). Control PHCs also pass constituency tests:4 they may be topicalized (7) or undergo contrastive dislocation (8) (Thráinsson 1979, 2007, Zaenen 1980, 1997, Ott 2014b).5

(7)

  • a.

    [Það að PROi heimsækja Ólaf] ákváðu þeiri, en . . .

    [it.ACC to visit Olaf] decided they but

    ‘They decided to visit Olaf, but . . .’

  • b.

    [Því að PROi hitta Maríu] gleymdi égi . . .

    [it.DAT to meet Mary] forgot I

    ‘I forgot to meet Mary . . . ’

(8)

  • a.

    [Það að PROi heimsækja Ólaf], (einmitt) það ákváðu þeiri.

    [it.ACC to visit Olaf] (exactly) it.ACC decided they

    ‘To visit Olaf, they decided to do just that.’

  • b.

    [Því að PROi heimsækja Maríu], (einmitt) þvígleymdi égi alveg.

    [it.DAT to visit Mary ] (exactly) it.DAT forgot I.NOM completely

    ‘To visit Mary, I completely forgot to do just that.’

There is even further reason, beyond these constituency tests, to believe that the presence of the pronoun does not necessarily indicate that extraposition has applied. According to Thráinsson (1979:221), (9b) involves extraposition, while (9a) does not. Sentences (10a–b), then, show that the pronoun itself can be topicalized only when the clause has extraposed: (10a), in which the pronoun is topicalized from the nonextraposed order of (9a), is ungrammatical, while (10b), in which the pronoun is topicalized from the extraposed order of (9b), is grammatical.

(9)

  • a.

    Ég veit [það að María er farin] alveg fyrir víst.

    I know [it that Mary is gone ] quite for sure

    ‘I know for sure that Mary is gone.’

  • b.

    Ég veit það alveg fyrir víst [að María er farin].

    I know it quite for sure [that Mary is gone ]

(10)

  • a.

    ?*Það veit ég [að María er farin] alveg fyrir víst.

    it know I [that Mary is gone] quite for sure

  • b.

    Það veit ég alveg fyrir víst [að María er farin].

    it know I quite for sure [that Mary is gone]

These facts receive a straightforward explanation: if extraposition does not apply, topicalization moves the entire PHC, yielding sentences like (6a)/(7). If extraposition moves the CP out of the PHC, then topicalization of the DP remnant yields (10b).

Assuming this much, it is straightforward to test whether PHCs are islands when extraposition does not take place; the sentences in (11) verify that they are.

(11)

  • a.

    Ég veit [það að Jón elskar Maríu ] alveg fyrir víst.

    I.NOM know [it that John.NOM loves Mary.ACC] quite for sure

    ‘I know that John loves Mary for sure.’

  • b.

    *Maríui veit ég [það að Jón elskar ti] alveg fyrir víst.

    Mary.ACC know I.NOM [it that John.NOM loves ] quite for sure

    Intended: ‘Mary I know that John loves for sure.’

Now, one could try to claim that in (11a), the clause in fact has undergone extraposition, and that the adverbial alveg fyrir víst ‘for sure’ is simply right-adjoined to vP after the clause has adjoined there. This, however, requires a new explanation for (10b), which follows straightforwardly from the assumption that extraposition has not taken place.

To take another example, it has been noted that extraction is possible out of clauses that have undergone right-node raising (RNR).

(12) [Which official]i did they say that Bob suspected and Frank proved [that Sally bribed ti]? (Postal 1998:146)

In Icelandic, too, it is possible to extract out of a RNR clause—as long as the clause is not a PHC (see (6e)), as illustrated in (13)).

(13)

  • a.

    Hverni elskar Jón en aðrir hata [(*það) að deildin vill ráða ti]?

    who.ACC loves John but others hate [(*it.ACC) that the.department wants hire ]

    ‘Who does John love but others hate that the department wants to hire?’

  • b.

    Hverni er Jón ánægður með en aðrir eru ósáttir við [(*það) að

    who.ACC is John happy with but others are unhappy with [(*it.ACC) that

    deildin vill ráða ti]?

    the.department wants hire ]

    ‘Who is John happy with but others unhappy with the department wanting to hire?’

    (Einar Freyr Sigurðsson, pers. comm.)

The strength of this argument depends on whether extraposition has taken place in examples like (6e). If we assume that the only parse of (6e) is one in which extraposition has taken place, then (13) does not say anything beyond the general fact that extraction is impossible out of object extraposed clauses. I know of no good reason to suppose that extraposition has taken place in (6e); (6e) simply fits in with all the other constituency tests showing that the pronoun and the CP form a constituent. Why should such a constituent not be able to undergo RNR? Moreover, even if extraposition is one possible parse of (6e), the argument goes through so long as it is not the only parse. Suppose there is one parse of (6e) where the pronoun and the CP form a surface constituent, and another where the CP has extraposed out of that constituent. If the islandhood of PHCs derives from extraposition, then extraction should be possible under the parse where extraposition has not taken place. But extraction is not possible at all, which follows from the claim that PHCs are islands whether extraposition takes place or not.6

So far, then, we have good reason to believe that for PHCs, the pronoun and the clause form a constituent, and that extraction out of the CP contained in this constituent is not possible, regardless of whether extraposition has taken place. In the next section, this point will be strengthened by an examination of subject extraposition.

3 Subject Extraposition

To appreciate the point about constituency discussed in the previous section, it is important to turn to subject extraposition. In short, subject extraposition shows that it is not extraposition that creates an extraction island per se; it is the PHC that is an island for extraction. That is, sometimes subject extraposition is movement of a CP out of a PHC, but sometimes it is just movement of a CP. Only in the former case is the extraposed clause an island. Importantly, Thráinsson (1979) argues that unlike subject extraposition, CP object extraposition with an overt pronoun is always movement out of a PHC. Thus, object extraposition is an extraction island not because of the landing site of extraposition—not even because of extraposition at all—but because object extraposition involves a PHC, and PHCs are extraction islands.

Thráinsson (1979) was investigating a well-known proposal by Rosenbaum (1967), according to which sentential arguments are NPs headed by a sometimes silent pronoun ‘it’. According to this view, extraposition involves movement of the CP (to update the terminology) out of this NP constituent. Thráinsson’s conclusion was that the Rosenbaum structures do exist—but that another option also exists, namely, extraposition of the CP, leaving behind a dummy pronoun. That is, it cannot be maintained that all CP-extraposition involves movement out of an NP.

Consider (14), which can be derived by extraposing the CP out of the PHC in (15a), or by extraposing the clause in (15b) and inserting a dummy expletive.

(14)

  • Það er merkilegt [að jörðin skuli vera hnöttótt].

  • it is interesting [that the.earth should be round ]

  • ‘It is interesting that the earth is round.’

  • (Thráinsson 1979:191)

(15)

  • a.

    [Það að jörðin skuli vera hnöttótt] er merkilegt.

    [it that the.earth should be round ] is interesting

    ‘That the earth is round is interesting.’

    (Thráinsson 1979:190)

  • b.

    [Að jörðin skuli vera hnöttótt] er merkilegt.

    [that the.earth should be round ] is interesting

    ‘That the earth is round is interesting.’

    (Thráinsson 1979:193)

The difference between the two derivations can be seen by forming a yes/no question from the sentence in (14). The dummy pronoun is not retained in such contexts. (16a) is unambiguously derived from a structure such as (15a), while (16b) is unambiguously derived from a structure such as (15b).

(16)

  • a.

    Er það (ekki) merkilegt [að jörðin skuli vera hnöttótt]?

    is it (not) interesting [that the.earth should be round ]

    ‘Is(n’t) it interesting that the earth is round?’

    (Thráinsson 1979:192)

  • b.

    Er (ekki) merkilegt [að jörðin skuli vera hnöttótt]?

    is (not) interesting [that the.earth should be round ]

    ‘Is(n’t) it interesting that the earth is round?’

    (Thráinsson 1979:194)

At least three further pieces of evidence corroborate this view. First, for many predicates, PHCs require a special discourse context. Thráinsson (1979) notes that with nonfactive and semifactive predicates such as satt ‘true’, líklegt ‘likely’, and augljóst ‘obvious’, the pronoun is often unacceptable without a special discourse context.

(17)

  • #[Það að Jón hefur étið hákarlinn] er augljóst.

  • [it that John has eaten the.shark] is obvious

  • (Thráinsson 1979:181)

If the proposition is mentioned in the discourse first (providing an antecedent for the pronoun), then such sentences are possible, as illustrated with the dialogue in (18).

(18)

  • A: Það er augljóst að María heldur við Harald og elskar Jón ekki neitt.

    ‘It’s obvious that Mary’s having an affair with Harold and doesn’t love John at all.’

    it is obvious that Mary holds with Harold and loves John not at.all

  • B: [Það að María heldur við Harald] er augljóst, (en ég gæti nú samt trúað

    [it that Mary holds with Harold] is obvious but I could now still believe að hún elskaði Jón).

    that she loved John

    ‘That Mary’s having an affair with Harold is obvious, but I could still believe that she loves John.’

    (Thráinsson 1979:202)

When the pronoun is retained in noninitial contexts and the CP extraposes, the same requirement for a special discourse context shows up.

(19)

  • #Nú er Það augljóst að María heldur við Harald.

    now is it obvious that Mary holds with Harold

    ‘Now it’s obvious that Mary’s having an affair with Harold.’

    (Thráinsson 1979:203)

Thus, the same kind of discourse context is required for (a) extraposition sentences where noninitial það is retained, and (b) nonextraposed sentences containing a PHC in the subject position. This correlation supports the view that when the pronoun is retained, it is retained because extraposition takes place out of the PHC.

Second, we saw in (1b) that fresta ‘postpone’ assigns dative to its DP object when there is one. As shown in (20), this pronoun may occupy the subject position when the verb is passivized, even in noninitial contexts.

(20)

  • Í gær var því frestað að hálshöggva fangana.

  • yesterday was it.DAT postponed to execute the.prisoners

  • ‘Yesterday, executing the prisoners was postponed.’

  • (Thráinsson 1979:229)

However, (1b) showed that the pronoun is optional with fresta ‘postpone’. Thus, it is also possible to have an extraposition sentence that does not have a dative pronoun in the subject position, as shown in (21a). However, since this pronoun is a dummy, it cannot be retained in noninitial contexts, as illustrated in (21b).

(21)

  • a.

    Það var frestað að hálshöggva fangana.

    it.NOM was postponed to execute the.prisoners

    ‘Executing the prisoners was postponed.’

    (Thráinsson 1979:228)

  • b.

    Í gær var (*það) frestað að hálshöggva fangana.

    yesterday was (*it.NOM) postponed to execute the.prisoners

    ‘Yesterday, executing the prisoners was postponed.’

    (Thráinsson 1979:229)

The source of the contrast between (21b) and (16a) is clear: the pronoun in (16a) is the same one that forms a constituent with the clause in (15a). When a noninitial pronoun is retained in extraposition contexts, it is the pronoun from the PHC. For fresta ‘postpone’, that pronoun is dative, so the nominative það in (21b) is not possible.

The third point is in fact a variant of the second point. For some verbs, the pronoun is not optional; it is obligatory. For example, fagna ‘rejoice’ requires a dative pronoun in order to take a clausal complement, as illustrated in (22a). When verbs with this property are passivized and their clauses extraposed, the pronoun in the subject position must be dative, not nominative, as shown by the contrast in (22b).

(22)

  • a.

    Ég fagna *(Því) að Þú skulir vera kominn.

    I rejoice *(it.DAT) that you should be arrived

    ‘I rejoice in your arrival.’

  • b.

    {Því/ *Það } er fagnað að Þú skulir vera kominn.

    {it.DAT / *it.NOM} is rejoiced that you should be arrived

    (Thráinsson 1979:230)

These facts reinforce the claim above that subject extraposition may take place out of a PHC, stranding the pronoun, or not, leading to a dummy pronoun. When the PHC pronoun is obligatory in the active, the dummy pronoun is impossible in the passive.

We have seen solid evidence that (a) pronouns may form a constituent with a CP clause, and (b) extraposition only sometimes moves a CP out of this clause; other times, the CP moves in its entirety and leaves a first-position dummy pronoun behind. Given this much, it is straightforward to show that extraction out of an extraposed CP is only banned when a PHC is involved.

(23)

  • a.

    Maríui er hörmulegt [að Jón skuli hafa barið ti].

    Mary.ACC is deplorable [that John should have hit ]

    ‘Mary, it’s deplorable that John hit.’

    (Thráinsson 1979:195)

  • b.

    *Maríui er Það hörmulegt [að Jón skuli hafa barið ti].

    Mary.ACC is it.NOM deplorable [that John should have hit ]

    (Thráinsson 1979:196)

In (23a), the object Maríu ‘Mary.ACC’ has been extracted out of the embedded clause and topicalized to the matrix CP. The extraction is successful, since the CP is not at any point contained within a PHC. The absence of a PHC is verified by the fact that there is no dummy pronoun; the dummy pronoun only appears in the first position, and since the DP Maríu ‘Mary.ACC’ has moved to Spec,CP, it is that DP that is in the first position, making the dummy pronoun impossible. In (23b), the pronoun is retained and the sentence is ungrammatical: the retention of the noninitial pronoun indicates extraposition out of a PHC, and PHCs are islands.

By adding one more clause layer, it can even be shown that the overt presence of Það is not the problem. In (24), the intermediate Það is the dummy pronoun, since it is in the first position (in its clause) and extraction is possible past it.

(24)

  • Maríui segir Ólafur að Það sé hörmulegt [að Jón skuli hafa barið ti].

  • Mary.ACC says Olaf.NOM that it is deplorable [that John should have hit ]

  • ‘Mary, Olaf says it’s deplorable that John hit.’

  • (Thráinsson 1979:197)

Thráinsson (1979:197) sums up his conclusion as follows: “Thus, it is only the Það which is unmistakably a remnant of a base generated [NP Það S] configuration which ‘blocks’ extraction and not just any Það nor a particular semantic class of predicates.”

What the facts discussed in this and the previous section show is that for the argument against the MTC, it does not matter when extraposition happens, or what projection it targets; the PHC is an extraction island whether extraposition takes place or not. Moving the controller out of the clause before extraposition takes place does not help anything, because the clause is already an island, before extraposition takes place. The fact that control is possible across this boundary shows that the dependency between PRO and its controller cannot be reduced to A-movement as we currently understand it; my objection in Wood 2012 therefore stands.

4 ‘Promise’-Type Verbs and Other Issues

The preceding sections have outlined why D&H’s response does not suffice to handle the facts I discussed in Wood 2012. In short, their response is based on a misunderstanding of the analysis of Icelandic object extraposition. However, D&H’s reply raises a number of further issues, two of which I will address briefly in this section. First, I present a novel argument against the MTC from Icelandic ‘promise’-type verbs;7 then I discuss a further problem stemming from raising-to-object constructions.

4.1 ‘Promise’-Type Verbs

D&H discuss ‘promise’-type verbs, which seem to be problematic for the MTC.

(25) Johni promised Maryj [PROi/*j to leave].

In (25), John controls PRO even though Mary seems to intervene, in violation of Rosenbaum’s (1967) Minimal Distance Principle. In response to Ndayiragije 2012, D&H cite the solution to this problem proposed in Hornstein 2001 and Boeckx and Hornstein 2003:274: namely, to suppose that Mary is contained in a PP headed by a null preposition. The ability of John to move past Mary with ‘promise’-type verbs is thus collapsed with the ability of John to move past the PP to Mary with raising verbs.

(26)

graphic

This explanation cannot extend to Icelandic ‘promise’-type verbs.8 In Icelandic, both the DP object of lofa ‘promise’ and the experiencer of virðast ‘seem’ are dative DPs. Unlike in English, however, the experiencer cannot be skipped by A-movement—it is the experiencer that undergoes A-movement to the subject position (Sigurðsson 1996, Boeckx 2000, Holmberg and Hróarsdóttir 2004, SigurLsson and Holmberg 2008).

(27)

  • a.

    Méri hefur alltaf virst ti [Jón vera klár ].

    me.DAT has always seemed [John.NOM be clever]

    ‘John has always seemed to me to be clever.’

  • b.

    *Jóni hefur alltaf virst mér [ti vera klár ].

    John.NOM has always seemed me.DAT [ be clever]

The “promisee” of lofa ‘promise’ is demonstrably unskippable as well. When it takes two DP objects, both objects are marked dative, as shown in (28a). In the passive, it is the higher object, which expresses the person receiving the promise, that undergoes A-movement to the subject position. This object may not be skipped by the lower object.9

(28)

  • a.

    Þeir hafa lofað bændunum peningunum.

    they have promised the.farmers.DAT the.money.DAT

    ‘They have promised the farmers the money.’

  • b.

    Bændunumi hefur verið lofað ti peningunum.

    the.farmers.DAT have been promised the.money.DAT

    ‘The farmers have been promised the money.’

    (Thráinsson 2007:240)

  • c.

    *Peningunumi hefur verið lofað bændunum ti.

    the.money.DAT has been promised the.farmers.DAT

    Intended: ‘The money has been promised to the farmers.’

Given that argumental dative DPs in general, and the object of lofa ‘promise’ in particular, cannot be skipped by A-movement, the MTC would lead us to expect that ‘promise’-type control is not possible in Icelandic. This expectation is not borne out; Icelandic behaves exactly like English. Either the subject can be the controller, as in (29a), or the object can be the controller, as in (29b); this is exactly like English, as indicated in the translations. Note also that a PHC is possible in both cases.

(29)

  • a.

    Jóni lofað i mérj [(Því) [að PROi berja mig ekki aftur ]].

    John.NOM promised me.DAT [(it.DAT) [to beat me not again]]

    ‘John promised me to not beat me again.’

  • b.

    Jóni lofaði mérj [(Því) [að PROj fá að fara bráðum]].

    John.NOM promised me.DAT [(it.DAT) [to get to leave soon ]]

    ‘John promised me to be allowed to leave soon.’

To derive (29a), the MTC would require A-movement of Jón past mér ‘me’ (as well as the pronoun when it is present), and there is no reason to believe that this is possible.

Finally, it is worth pointing out that exactly like passives in English, passives in Icelandic disambiguate the object and subject control readings. The implicit agent cannot control PRO when the object moves to the subject position.10

(30)

  • a.

    *Méri var lofað [(Því) [að PROj berja mig ekki aftur ]].

    me.DAT was promised [(it.DAT) [to beat me not again]]

    Like English: ‘*I was promised not to beat me again.’

  • b.

    Méri var lofað [(Því) [að PROi fá að fara bráðum]].

    me.DAT was promised [(it.DAT) [to get to leave soon ]]

    ‘I was promised to be allowed to leave soon.’

That is, there is no relevant difference between Icelandic and English ‘promise’-type verbs except that (a) dative case is retained (a general difference between passives of indirect objects in Icelandic and English), (b) a PHC is possible (a general difference between Icelandic and English control infinitives), and (c) the explanation for ‘promise’-type verbs given by Hornstein (2001), Boeckx and Hornstein (2003), and D&H does not work for Icelandic.

4.2 Raising to Object and Extraposition

D&H’s reply raises a number of further issues. D&H (p. 461) note that their account is not easily reconcilable with the view that Ā-extraction proceeds through vP edges, as required by the theory of phases in Chomsky 2001.11 It would also seem to be incompatible with Chomsky’s (2008) analysis of ECM (exceptional case marking) as raising to object. In brief, Chomsky (2008) proposes that in ECM constructions, V inherits ɸ-features from v* and enters an Agree relation with the embedded subject, which then remerges in the matrix Spec,VP. In fact, there is long-standing empirical evidence in favor of the claim that ECM constructions involve raising to object, going back at least to Postal 1974, and Thráinsson (1979) argues that this is correct for Icelandic as well. If the reason that raising infinitives cannot be extraposed is that they involve A-movement higher than vP, as D&H propose, then we might expect that extraposition is possible in raising-to-object constructions. After all, when the embedded subject moves to Spec,VP, the landing site is even lower than in the MTC derivation of control clauses. So there should be no problem with extraposing the embedded clause after that. As shown in (31), this is not possible (even though object control into a PHC is possible, as shown in (29b)).

(31)

  • Hanni hefur alltaf talið mig íbarnaskap sínumi (*Það) vera prest.

  • he has always believed me in foolishness REFL (*it.ACC) be priest

  • ‘He has always believed me in his foolishness to be a priest.’

Of course, it is always possible (perhaps even likely) that extraposition of raising infinitives is ungrammatical for independent reasons. Perhaps raising infinitives are TPs and control infinitives are CPs, and TPs, for some reason, cannot extrapose. However, this is not what D&H say. They explicitly rule out extraposition of raising infinitives on the grounds that raising targets a particular region of the clause—Spec,TP—that is higher than the vP target of extraposition. That explanation does not extend to raising-to-object constructions.12

5 Conclusion

In this article, I have shown that object extraposition remains a problem for the MTC. As argued by Thráinsson (1979), object extraposition is derived from a pronoun-headed clause (PHC), a structure where a pronoun forms a constituent with a clause. Movement out of a PHC is impossible regardless of whether or not extraposition takes place. D&H’s attempt to account for the Icelandic facts on the basis of the landing site of extraposition cannot succeed, because the PHC is an island before extraposition takes place, and even if it does not take place. Even though movement dependencies cannot reach inside PHCs, control dependencies can. The conclusion is that the locality for control is not the same as the locality for movement.

I have furthermore shown that D&H’s reply to Ndayiragije 2012 cannot be extended to Icelandic ‘promise’-type verbs. If we take the locality of A-movement seriously—something that is well-understood for Icelandic—then there is no explanation for why subject control is possible past an object with ‘promise’-type verbs. The object in question is an ordinary dative DP that undergoes ordinary A-movement, and blocks A-movement of DPs past it. If we truly want to understand the various dependencies encoded in natural language, we cannot conclude that control and A-movement dependencies are collapsed. Their locality properties are just not the same.

Notes

I would like to thank the anonymous LI reviewers for helpful comments and suggestions on this article, as well as Anton Karl Ingason, Halldór Ármann Sigurðsson, Neil Myler, Dennis Ott, and Einar Freyr Sigurðsson (along with anyone I might have forgotten) for numerous helpful discussions. Thanks to Anton Karl Ingason, Einar Freyr Sigurðsson, and Höskuldur Thráisson for native-speaker judgments of Icelandic sentences. I am responsible for any remaining errors.

1 The pronoun has the following forms: það ‘nominative/accusative’ (syncretic), því ‘dative’, þess ‘genitive’.

2 Note, moreover, that according to Thráisson (1979:279), a pronoun referring to an infinitive clause is in fact possible, paceOtt (2014a:61).

  • (i)

    • A: Hætti Jón að reykja?

      stopped John to smoke

      ‘Did John stop smoking?’

    • B: Já hann hætti því.

      yes he stopped it.DAT

      ‘Yes, he stopped.’

3Thráisson (1979) uses the notation [NP það S] for his claim; I will assume that the constituent is [DP það CP] (see also Fischer 2013), but nothing hinges on this.

4 Thanks to Höskuldur Thráinsson and Anton Karl Ingason for judgments of these examples. Anton Karl Ingason finds the contrastive dislocation examples better with a focus particle like einmitt ‘exactly’, and Höskuldur Thráinsson emphasizes that the pronoun must be contrastively focused in this construction. In some cases, the Icelandic sentences do not translate naturally into English sentences with similar fronting operations, so I opt for a translation without such fronting. The point, of course, is what Icelandic fronting shows about constituency.

5 Some constituency tests, such as passivization, do not apply well to control PHCs owing to the anaphoric nature of PRO. See Thráinsson 1979:117–118 for discussion. I will take it that the examples in (7) and (8) are enough to show that control clauses headed by a pronoun do form a constituent with that pronoun, just as the finite clauses do, and that extraposition is movement of the CP out of that constituent.

6Ott (2014a) claims that extraposed infinitives are not a problem for the MTC under an analysis that treats extraposition as right-dislocation, and a biclausal analysis of right-dislocation at that. First of all, since Thráisson (1979) carefully distinguishes between these constructions and right-dislocation constructions, and they have distinct prosodic and syntactic properties, Ott’s claim is almost certainly not correct in the general case. Second, (9a) would seem to be a counterexample to any proposal linking the pronoun to its clause by right-dislocation only, assuming that the adverbial alveg fyrir víst ‘for sure’ (lit. ‘quite for sure’) is attached to the matrix clause. (Attaching the adverbial to the mostly deleted clause of the right dislocate might necessitate nonconstituent deletion, or unmotivated movement operations, for example.) Finally, Ott’s (2014a:60) statement that his argument “holds as long as the Right-dislocation parse is available,” given a situation where the sentences are “derivationally ambiguous,” is incorrect. It is the other way around: if there is a parse that cannot be analyzed as right-dislocation—or even as object extraposition at all—and extraction is not possible while control is, then the argument against the MTC goes through.

7 The relevance of ‘promise’-type verbs to the MTC has in general been widely discussed, as the references in the text show, but Icelandic ‘promise’-type verbs have not, to my knowledge, figured in this discussion.

8 Thanks to Einar Freyr Sigurðsson for discussion of the facts in this section.

10 At first glance, it would seem as though these facts are counterexamples to Van Urk’s (2013) claim (based on Landau’s (2000) system) that effects like this are limited to cases where T agrees with an overt DP. In (30), the subject is dative, so T has default agreement. However, there are actually reasons to think that T does abstractly agree with nonnominative subjects in Icelandic (in at least some features), and this has been proposed numerous times, for example by Boeckx (2000), Anagnostopoulou (2003), Schütze (2003), Koopman (2006), and Sigurðsson and Holmberg (2008). Under this hypothesis, the Icelandic facts are exactly as Van Urk (2013) would expect them to be.

11 See Van Urk and Richards 2015 for evidence that DPs undergoing Ā-movement do move through vP edges.

12 The point of invoking raising infinitives in the first place was simply that there are no contexts where we see positive evidence of A-movement out of a PHC; raising infinitives are closer than any other structures to having the properties where we might expect such movement to be possible. A reviewer, for example, suggests that if raising clauses are TPs and control clauses CPs, perhaps the PHC pronoun can only combine with CPs. This is entirely possible; but the point here is that we find no cases where anything can undergo A- or Ā-movement out of a PHC, and a control dependency into the PHC is nevertheless systematically allowed.

References

Anagnostopoulou,
Elena
.
2003
.
The syntax of ditransitives: Evidence from clitics
.
Berlin
:
Walter de Gruyter
.
Boeckx,
Cedric
.
2000
.
Quirky agreement
.
Studia Linguistica
54
:
354
380
.
Boeckx,
Cedric
, and
Norbert
Hornstein
.
2003
.
Reply to “Control is not movement.”
Linguistic Inquiry
34
:
269
280
.
Boeckx,
Cedric
, and
Norbert
Hornstein
.
2006
.
Control in Icelandic and theories of control
.
Linguistic Inquiry
37
:
591
606
.
Boeckx,
Cedric
,
Norbert
Hornstein
, and
Jairo
Nunes
.
2010a
.
Control as movement
.
Cambridge
:
Cambridge University Press
.
Boeckx,
Cedric
,
Norbert
Hornstein
, and
Jairo
Nunes
.
2010b
.
Icelandic control really is A-movement: Reply to Bobaljik and Landau
.
Linguistic Inquiry
41
:
111
130
.
Chomsky,
Noam
.
1995
.
The Minimalist Program
.
Cambridge, MA
:
MIT Press
.
Chomsky,
Noam
.
2001
. Derivation by phase. In
Ken Hale: A life in language
, ed. by
Michael
Kenstowicz
,
1
52
.
Cambridge, MA
:
MIT Press
.
Chomsky,
Noam
.
2008
. On phases. In
Foundational issues in linguistic theory: Essays in honor of Jean-Roger Vergnaud
, ed. by
Robert
Freidin
,
Carlos P.
Otero
, and
Maria Luisa
Zubizarreta
,
133
166
.
Cambridge, MA
:
MIT Press
.
Drummond,
Alex
, and
Norbert
Hornstein
.
2014
.
Some purported problems for the movement theory of control
.
Linguistic Inquiry
45
:
449
462
.
Fischer,
Silke
.
2013
.
Object extraposition in Germanic – A challenge for control theory?
Paper presented at the 28th Comparative Germanic Syntax Workshop, University of Leipzig
.
Holmberg,
Anders
, and
Thorbjörg
Hróarsdóttir
.
2004
.
Agreement and movement in Icelandic raising constructions
.
Lingua
114
:
651
673
.
Hornstein,
Norbert
.
1999
.
Movement and control
.
Linguistic Inquiry
30
:
69
96
.
Hornstein,
Norbert
.
2001
.
Move! A Minimalist theory of construal
.
Oxford
:
Blackwell
.
Koopman,
Hilda
.
2006
. Agreement configurations: In defense of “Spec head.” In
Agreement systems
, ed. by
Cedric
Boeckx
,
159
199
.
Philadelphia
:
John Benjamins
.
Landau,
Idan
.
2000
.
Elements of control: Structure and meaning in infinitival constructions
.
Dordrecht
:
Kluwer
.
Ndayiragije,
Juvénal
.
2012
.
On raising out of control
.
Linguistic Inquiry
43
:
275
299
.
Ott,
Dennis
.
2014a
.
Controlling for movement: Reply to Wood 2012
.
Working Papers in Scandinavian Syntax
92
:
58
65
.
Ott,
Dennis
.
2014b
.
An ellipsis approach to contrastive left-dislocation
.
Linguistic Inquiry
45
:
269
303
.
Postal,
Paul M
.
1974
.
On raising: One rule of English grammar and its theoretical implications
.
Cambridge, MA
:
MIT Press
.
Postal,
Paul M
.
1998
.
Three investigations of extraction
.
Cambridge, MA
:
MIT Press
.
Rosenbaum,
Peter
.
1967
.
The grammar of English predicate complement constructions
.
Cambridge, MA
:
MIT Press
.
Schütze,
Carson T
.
2003
. Syncretism and double agreement with Icelandic nominative objects. In
Grammatik i fokus/Grammar in focus: Festschrift for Christer Platzack 18 November 2003
, ed. by
Lars-Olof
Delsing
,
Cecilia
Falk
,
Gunlög
Josefsson
, and
Halldór Ármann
Sigurðsson
,
2
:
295
303
.
Lund
:
Lund University, Department of Scandinavian Languages
.
Sigurðsson,
Halldór Ármann
.
1989
.
Verbal syntax and case in Icelandic
.
Doctoral dissertation, Lund University
.
Sigurðsson,
Halldór Ármann
.
1996
.
Icelandic finite verb agreement
.
Working Papers in Scandinavian Syntax
57
:
1
46
.
Sigurðsson,
Halldór Ármann
, and
Anders
Holmberg
.
2008
. Icelandic dative intervention: Person and number are separate probes. In
Agreement restrictions
, ed. by
Roberta
D’Alessandro
,
Susan
Fischer
, and
Gunnar Hrafn
Hrafnbjargarson
,
251
280
.
Berlin
:
Mouton de Gruyter
.
Thráinsson,
Höskuldur
.
1979
.
On complementation in Icelandic
.
Doctoral dissertation, Harvard University, Cambridge, MA
.
Thráinsson,
Höskuldur
.
2007
.
The syntax of Icelandic
.
Cambridge
:
Cambridge University Press
.
van Urk,
Coppe
.
2013
.
Visser’s Generalization: The syntax of control and the passive
.
Linguistic Inquiry
44
:
168
178
.
van Urk,
Coppe
, and
Norvin
Richards
.
2015
.
Two components of long-distance extraction: Successive cyclicity in Dinka
.
Linguistic Inquiry
46
:
113
155
.
Wood,
Jim
.
2012
.
Against the “movement theory of control”: Another argument from Icelandic
.
Linguistic Inquiry
43
:
322
330
.
Wood,
Jim
, and
Halldór Ármann
Sigurðsson
.
2014
.
Let-causatives and (a)symmetric DAT-NOM constructions
.
Syntax
17
:
269
298
.
Zaenen,
Annie
.
1980
.
Extraction rules in Icelandic
.
Doctoral dissertation, Harvard University, Cambridge, MA
.
Zaenen,
Annie
.
1997
. Contrastive dislocation in Dutch and Icelandic. In
Materials on left dislocation
, ed. by
Elena
Anagnostopoulou
,
Henk
van Riemsdijk
, and
Frans
Zwarts
,
119
148
.
Amsterdam
:
John Benjamins
.
Zaenen,
Annie
,
Joan
Maling
, and
Höskuldur
Thráinsson
.
1985
.
Case and grammatical functions: The Icelandic passive
.
Natural Language and Linguistic Theory
3
:
441
483
.