2 Objects of promise verbs in Russian are not prepositional

The MTC-compliant analysis of subject control for promise-type verbs schematised in (2) is predicated on the reality of the posited null preposition as a syntactic object. The preposition is crucial as without it, the object position of promise would have been the closest position targeted by A-movement, resulting in obligatory object control, contrary to fact. If the object of promise is introduced by a covert preposition, then the closest position targeted by A-movement is the subject position, yielding obligatory subject control.

Boeckx et al. (Reference Boeckx, Hornstein and Nunes2010a) diagnose the presence of this silent preposition in English by noting a number of asymmetries between the objects of promise on the one hand, and those of, for instance, order on the other. Because the analysis in (2) is claimed to be universal, the onus is on the MTC to show that identical or at least comparable asymmetries hold between the objects of promise and order crosslinguistically. We now show that this is not the case in Russian.Footnote ^[2]

The first asymmetry concerns the accessibility of the object argument for wh-extraction, as shown in the examples below:

While in English only the objects of order-type verbs can undergo wh-movement, Russian objects of promise-type verbs are equally accessible:Footnote ^[3]

Similarly, only the objects of order-type verbs can undergo heavy NP shift in English, whereas those of promise-type verbs disallow it:

In Russian, however, the objects of both types of verb can undergo heavy NP shift, as illustrated in (8) below, where a dative-marked object modified by a relative clause appears right-peripherally.

Therefore, while the contrasts reported by Boeckx et al. (Reference Boeckx, Hornstein and Nunes2010a) could be taken as evidence of a structural difference between the objects of order-type verbs on the one hand and the objects of promise-type verbs on the other, there is no evidence of such a structural difference between their Russian counterparts. As just shown, objects of verbs such as prikazat’ ‘order’ and poobeshchat’ ‘promise’ in Russian share both morphosyntactic and syntactic properties: they receive identical case marking and are equally accessible for wh-movement and heavy NP shift. Postulating a silent preposition governing some of these objects but not the others is unmotivated.

Because the absence of evidence is not necessarily evidence of absence, we now present a morphophonological argument against positing a silent preposition governing the objects of promise-type verbs coming from pronoun n-allomorphy. Russian third-person pronouns appear in the so-called ‘j-form’ when they are not complements of prepositions (Hill Reference Hill1977, Chvany Reference Chvany1982, Timberlake Reference Timberlake2004). When governed by a preposition, however, they surface in the so-called ‘n-form’. These forms are not in free variation: the ‘j-form’ cannot appear in the complement of a preposition, nor can the ‘n-form’ appear when it is not governed by a preposition.

As example (10) illustrates, when the object of a promise-type verb is a third-person pronoun, that pronoun can only take the non-prepositional ‘j-form’ im, while the prepositional ‘n-form’ nim is unacceptable.

It may be suggested that the ‘n-form’ is obligatory for complements of overt prepositions, in which case the pronominal object in (10) could be argued to be governed by a covert preposition that need not enforce the ‘n-form’. To see that this is wrong, let us consider the only other context in Russian where the ‘n-form’ is attested: phrasal comparatives.

Philippova (Reference Philippova2018: Chapter 5) argues convincingly that the ‘n-form’ of third-person pronouns in Russian is licensed locally by a null preposition introducing the standard in phrasal comparatives. Thus, null prepositions, just like their overt counterparts, require that their third-person pronominal complements appear in the ‘n-form’. Consequently, the resistance of the object in (10), which is purportedly governed by a covert preposition, to taking on the ‘n-form’ remains unaccounted for. We conclude that in Russian, the objects of promise-type verbs are not introduced by covert prepositions, making the analysis of subject control in (2) an ill fit for Russian.

3 Russian oblique control is obligatory control

As noted by Landau (Reference Landau, Davies and Dubinsky2007: fn. 5), the analysis of subject control sketched in (2) for promise-type verbs makes the strong prediction that complements of overt prepositions should be unable to participate in obligatory control. This prediction is incorrect, as evidenced by the existence of oblique control:

While the very existence of oblique control is a problem for the MTC since A-movement is known not to target oblique (e.g. complement-of-P) positions, there are two ways of rendering oblique control unproblematic for the MTC. One is to deny that examples like (12) instantiate obligatory control, thereby removing it from the purview of the MTC. The other is to appeal to sidewards movement to have the controller legitimately move into a non-commanding position. We argue in this section and the next that neither solution is adequate.

We begin by showing that oblique-control structures in Russian involve an unpronounced subject contained in the infinitival clause; the infinitival clause is thus not subjectless. The relevant evidence comes from the behaviour of reflexive pronouns, which in Russian are subject-oriented (Rappaport Reference Rappaport1986: 101). If the infinitival clause embedded under an oblique-control verb such as potrebovat’ ‘demand’ contains a reflexive pronoun, as in (13) below, the reflexive can be anteceded, descriptively speaking, by the oblique controller, which suggests the infinitival clause contains a hidden subject binding the reflexive.

We are ready to show, using conventional diagnostics, that the relation between the oblique controller and the hidden subject is that of obligatory control.

First, the prepositional object is the only nominal in the sentence that can bind the understood subject of the infinitival clause.

In (14), it is only the prepositional object brat Vasi ‘Vasya’s brother’ and not Serëzha, Lena or Vasya himself that can be construed as PRO.

Second, only the sloppy reading is possible under ellipsis in the constructions at hand so that for a sentence such as (15), the strict reading is unavailable.

Finally, if the oblique controller is modified by the focus-sensitive particle tol’ko ‘only’, the silent subject of the infinitive clause is obligatorily bound by the oblique controller as in (16), evidenced by the fact that the alternatives with respect to which the semantics of the entire sentence is computed have the general form ‘ $\unicode[STIX]{x1D706}x$ . The chief demanded from $x$ that $x$ hand in the report’, where the subject of the infinitive clause covaries with the matrix addressee argument.

We conclude that, from the point of view of traditional diagnostics, oblique control in Russian instantiates obligatory, rather than non-obligatory, control.

4 An unviable analytic option: Sidewards movement

One of the main problems with deriving oblique control by means of A-movement is that the movement would have to target a non-commanding position (i.e. the complement of an overt preposition). A way of sidestepping this issue is to employ Nunes’s (Reference Nunes2004) sidewards movement, which may target non-c-commanding positions.

According to Nunes (Reference Nunes2004), every moving expression must be internally merged to the root, but not necessarily that of the tree where the expression in question originated. That is, if there are two syntactic objects, A and B, in our workspace, an expression $\unicode[STIX]{x1D6FC}$ contained within object A may be copied and internally merged to the root of object B, turning object B into object C, as illustrated in (17). After that, objects A and C may be merged together.

While sidewards movement has been used to formalise adjunct control in terms of movement (Hornstein Reference Hornstein2001, Green Reference Green2019), we argue that it fails as an adequate account of oblique control in Russian. We use sentence (12) from above, repeated here as (18), as an illustration.

Let us suppose that object A is the infinitival clause containing the controller (Vasya) and object B is the preposition ot ‘from’. First, the controller is copied from the infinitival clause and merged with the prepositional head, thereby creating the oblique controller PP, as in (19b). Then the matrix verb merges with the infinitival clause, whereupon the prepositional phrase is inserted in the right structural position in the matrix clause (e.g. as a specifier of an Appl head), as in (19c).

However, the resulting structure is not linearisable: for the lowest copy of Vasya to be deleted, the highest copy must asymmetrically c-command all the lower copies, which is normally achieved by having the sidewards-moved element undergo another movement step following its integration into the host syntactic structure. This movement step is crucial since the only motivation behind the deletion of the lower copies, according to Nunes (Reference Nunes2004), is to stop them precluding linearisation. Building on Kayne’s (Reference Kayne1994) Linear Correspondence Axiom, Nunes assumes that two copies of the same element, one of which asymmetrically c-commands the other, will inevitably prevent linearisation since the expression will have to linearly precede itself. Therefore, all copies except one must be deleted for the purposes of linearisation.

With oblique control, however, there is neither any motivation for the oblique controller to vacate the PP after the prepositional phrase has been integrated into the matrix clause nor a target position to move to. Moreover, complements of Russian prepositions are unable to move, there being a strong ban on preposition stranding (Abels Reference Abels2003). Finally, even if such a movement were possible, the controller noun phrase would be pronounced not in the complement of the prepositional phrase, but in this mysterious position, since in Russian, it is usually the highest copy that is pronounced. Because the oblique controller is pronounced in the complement of the prepositional head, we can reliably assume that no movement of that noun phrase takes place after it is merged with the preposition. Consequently, neither copy of the controller noun phrase (one in the complement of the prepositional phrase and one in the subject position of the infinitival clause, say Spec,TP) asymmetrically c-commands the other. Then, they do not prevent the linearisation of the structure, and since unnecessary operations are costly and therefore are not executed by the computational system, neither of the two copies will be deleted and both will be pronounced.Footnote ^[4]

This prediction contradicts our observations: the controller noun phrase is pronounced only once, in the complement of the preposition, and the variants with multiple pronounced copies are unacceptable. Oblique control, therefore, does not follow from sidewards movement.

5 Oblique controllers do not originate inside infinitival clauses

In this section, we attempt to locate the lower copy of the putative A-movement chain, which the MTC would posit for oblique control. Because accusative object control is the closest parallel to oblique control that has been analysed as involving movement, the expectation is that oblique control and accusative object control should behave similarly with respect to the syntactic and semantic activities of the lower copy. We show, however, by examining the patterns of polarity licensing, case marking on floating quantifiers and the scope of numerical quantifiers, that oblique control systematically differs from object control in never revealing any activity of the purported lower copy and conclude that oblique control cannot be the result of movement. At the same time, we refrain from issuing a judgement on the nature of subject and object control.

5.1 Polarity licensing

Russian being a strict negative-concord language (Szabolcsi Reference Szabolcsi2018), all neg-words must be licensed by a predicate-mate negation, realised as ne, sometimes across a non-finite clause boundary (Timberlake Reference Timberlake2004: 259). Whilst negative indefinites are unacceptable in the absence of negation, as shown in (20), they need not follow the negation marker but can instead linearly precede it, as in (21) involving a negative indefinite in preverbal subject position.

Applying our heuristic for identifying movement dependencies, a negative indefinite originating in a negated infinitival clause is predicted to be licit even if it subsequently raises into the matrix clause.Footnote ^[5] In the case of oblique control, however, this expectation is not met: the controller in the matrix clause cannot be a negative indefinite licensed by predicate negation in the infinitival clause, as shown in (22a) and (22b).

The licit case of object control in (22c) shows that we are not simply dealing with a general prohibition on reconstructing A-movement. In this regard, examples (22b) and (22c) are particularly telling. The verb poprosit’ ‘ask’ in Russian allows the addressee argument (which controls PRO in the infinitival clause in both cases) to be realised either as a direct object (22c) or an oblique genitive nominal embedded inside a PP (22b). In the absence of any semantic differences between the two subcategorisation frames, a purely structural difference concerns the possibility of a negative indefinite controller being licensed within the infinitival clause, which is grammatical under object control (22c) but inadmissible under oblique control (22b).Footnote ^[6]

The same asymmetry between object control and oblique control is obtained with polarity items such as the nibud’-series of indefinite pronouns, whose acceptability depends on the presence of an operator – such as the imperative operator in (23b) – in the sentence (Haspelmath Reference Haspelmath1997, Pereltsvaig Reference Pereltsvaig, Halloway King and Sekerina2000, Reference Pereltsvaig, Abner and Bishop2008).

Being unquestionably acceptable as direct-object controllers, as in (24c), dependent indefinites of the nibud’-series are nevertheless poor oblique controllers:Footnote ^[7]

Once again, we see from the contrasts above that the controller DP in oblique-control environments cannot be licensed by an operator contained in the infinitival clause whereas such licensing is unproblematic in object-control environments.

5.2 Scope

Example (25) demonstrates that numerical quantifiers inside oblique controllers obligatorily take wide scope over the entire sentence rather than only the embedded infinitival clause, which is unexpected if oblique controllers originate in the infinitival clause.

In particular, the only available reading for (25) involves the existence of two specific employees to whom the appeal in question is addressed. The narrow-scope interpretation, whereby any two employees writing the report would satisfy the demand, is unavailable.

This contrasts with object control, which allows numerical quantifiers to take narrow scope:

The contrast between object control and oblique control with respect to allowing numerical quantifiers to scope low becomes especially salient when the numeral in question is accompanied by lyuboĭ ‘any’ with a view to forcing the narrow-scope interpretation:

Having identified robust differences between object control and oblique control with regard to polarity licensing and quantifier scope, we now show that morphosyntactic evidence reveals the same distinctions, reïnforcing our preliminary conclusion that, whilst subject and object control might at least be compatible with the MTC, oblique control clearly is not.

5.3 Case on floating quantifiers

Under oblique control in Russian, floating quantifiers in the embedded infinitival clause associated with its (PRO-)subject must take dative case, as shown in (29) below, whilst the oblique controller receives genitive case from the preposition.

The option of case transmission from the oblique controller in the matrix clause onto the floating quantifier in the embedded clause is unavailable, which stands in stark contrast to its general availability under object control, as in (30), and its obligatory character under subject control, as in (31). The contrast is unexpected since, on the MTC, case transmission should be the default pattern, all deviations from it requiring additional assumptions.Footnote ^[8]

The MTC analysis of case discord between the controller and the floating quantifier based on Icelandic data (e.g. Boeckx, Hornstein & Nunes Reference Boeckx, Hornstein and Nunes2010b) does not extend to Russian. It allows only two situations where a floating quantifier or a secondary predicate (FQ or SP) inside the infinitival clause could have a different case value from that of the controller in the matrix clause. The first is when the embedded infinitival verb is a quirky Case predicate and the FQ/SP bears this quirky Case. For the MTC, the controller originates inside the embedded clause, where it is assigned a theta role and gets its Case feature valued by the quirky Case predicate and where it agrees in Case (as well as $\unicode[STIX]{x1D711}$ -features) with the FQ/SP. The controller subsequently moves into the matrix clause, where it is assigned another $\unicode[STIX]{x1D703}$ -role from the matrix predicate and its Case feature value is overwritten by the one assigned in the matrix clause. This will not work for either oblique or object control in Russian since the dative case on FQs in (29) and (30) is not associated with any particular lexical verb or theta role (Comrie Reference Comrie, Brecht and Chvany1974).

The other MTC-compatible way to derive the Case mismatch between the controller and the FQ/SP in the infinitival clause, which is employed when the embedded predicate does not assign a quirky Case to its subject, is to have a default Case value on the FQ/SP (cf. Schütze Reference Schütze2001), which happens to be nominative in Icelandic. In Russian, however, the dative case in question has no properties traditionally ascribed to the default Case: in all the contexts of default Case assignment listed for English in Schütze (Reference Schütze2001), in Russian only a nominative noun phrase or no (overt) noun phrase is possible, but crucially not a dative one.

We conclude, on the basis of semantic and morphosyntactic evidence, that the oblique controller does not originate in the embedded infinitival clause.

6 Discussion and conclusions

In this paper, we have scrutinised an extension of the Movement Theory of Control, formulated by Boeckx et al. (Reference Boeckx, Hornstein and Nunes2010a), designed to account for the control asymmetry between promise-type verbs and order-type verbs, whereby the promise-type verbs require subject control as though the purported A-movement could bypass the object position. The extension in question posits a null preposition introducing the controller, thus making its complement position an unsuitable target for movement. Consequently, in order to confirm the MTC, its proponents must demonstrate the crosslinguistic reality of differences in argument structure between promise and order in each and every language, in dialect after dialect that have these constructions.

We have shown in the preceding sections, by appealing to a variety of sources of evidence, that in Russian, the movement-cum-silent-preposition view does not instantiate an adequate analysis of subject control for promise-type verbs: dative-marked objects of obeshchat’ ‘promise’ are syntactically indistinguishable from the dative-marked objects of prikazat’ ‘order’, nor is either of them introduced by a covert preposition.

For oblique control, we have argued that it instantiates obligatory control and should thus fall within the explanatory scope of the MTC. We have subsequently demonstrated the implausibility of movement analyses by showing that the oblique controller could not have moved into the complement of an overt preposition by sidewards movement nor originated inside the infinitival clause. We conclude that the MTC analysis of subject control in the case of promise-type verbs is untenable, reducing the MTC’s empirical coverage at most to regular subject and object control (and, when supplemented by sidewards movement, possibly also adjunct control).

As things stand, the facts as described in this squib are compatible with both a uniform non-movement PRO-based theory of control such as Landau (Reference Landau2015) and a hybrid theory such as Sheehan (Reference Sheehan, Huang, Poole and Rysling2014) or Grano (Reference Grano2015), whereby a subset of instances of obligatory control is derived via A-movement, the remaining ones being mediated by PRO.