1 Two Analyses of Partial Control
In partial control (PC), the null subject of a (typically) nonfinite complement is understood to be a group consisting of the referent of the matrix controller plus some other contextually salient individual(s) (Wilkinson 1971, Landau 2000, 2013).1
(Wei knew that) Harryj preferred [PROi+j to work together on the project].
(Elainei told Paulj that) shei would like [PROi+j to meet on Thursday].
On one analysis, PC PRO is syntactically singular but semantically plural at LF, like standard collective nouns (e.g., team, committee). A silent operator, licensed by certain matrix verbs, expands the reference of PRO to obtain the group reference (Pearson 2015, Landau 2016). On an alternative analysis, PC PRO is both syntactically and semantically singular. The PC reading arises indirectly from a null comitative phrase inside the complement (Hornstein 2003, Słodowicz 2008, Boeckx, Hornstein, and Nunes 2010:185, and specifically for Romance languages, Sheehan 2012, 2014). The two analyses are depicted in (2a) and (2b), respectively.2
Elainei preferred [PROi+ to meet on Thursday].
Elainei preferred [PROi to meet
with himon Thursday].
The two analyses differ on whether or not they take PC PRO to be semantically plural and on whether or not they posit a null comitative phrase. Therefore, grammatical tests that are sensitive to these properties can choose between them. Indeed, in earlier work (Landau 2007) I have pointed out that the null comitative analysis (barring stipulations) overgenerates nonexisting PC readings in simple clauses (3a) and undergenerates existing PC readings in complements containing collective predicates that do not take comitative phrases (3c) (see also the arguments in Sheehan 2014 against the null comitative analysis for English).
with himon Thursday.
*The chair dispersed with the rest of us.
The chair decided to disperse until next week.
In fact, simple reciprocal verbs in English resist a comitative phrase but can still appear in a PC complement.
*?Elaine kissed/hugged with Paul.
Paul felt that Elaine wanted to kiss/hug.
The question remains, however, whether the null comitative analysis is a viable account of sentences like (2b). Three novel arguments are presented here to show that it is not.
First, genuine collective nouns cannot bind singular personal reflexives, even when the intended referent of the latter is included in the group denoted by the noun (5a). Individual subjects of sentences with comitative phrases, however, may perfectly bind such emphatic reflexives (5b).
The team—that is, Peter’si team—met on Thursday (*himselfi).
Peteri met with them on Thursday himselfi.
PC PRO patterns with the former and not with the latter. This is unexpected if PC PRO is semantically singular and the plural reading arises from a null comitative phrase (note that (6) has an irrelevant reading where the emphatic reflexive is construed upstairs).
(6) Peter would like [PRO to meet on Thursday (*himself )].
Second, if PC PRO is semantically singular (as the null comitative analysis has it), it should be able to saturate a secondary predicate whose domain is restricted to nonplural individuals. Consider, for example, predicates like as a free man.
Peter will meet with Elaine as a free man tomorrow.
*Peter and Elaine will meet as a free man tomorrow.
*This couple will meet as a free man tomorrow.
(8) *Peter told Elaine that he expected to meet as a free man the following day.
Third, a singular subject with a discontinuous comitative phrase is not semantically equivalent to a group-denoting subject; its interpretation is, in fact, more specific (see Dimitriadis 2004, 2008, Siloni 2008, 2012). Consider the distribution of the adverb separately (not discussed in the works just mentioned).
Mary is meeting separately with the chair and the dean.
*Mary’s department is meeting separately.
Mary has been talking to the chair and the dean separately.
*Mary’s department has been talking to the chair separately.
One reading of separately, presently irrelevant, is discourse-anaphoric (the meeting was held separately from some other meeting, previously mentioned). The reading of interest here is the NP-dependent one. In (9a,c), it amounts to the interpretation that Mary had one meeting/talking event with the chair and another (different) meeting/talking event with the dean. On this reading, separately distributes over a set of events, each associated with a member of the plural NP in the comitative phrase. Since this reading requires some syntactically plural NP to individuate the events distributed over, it is not sanctioned by a collective noun like department in (9b,d), nor by a singular NP in the comitative phrase, like the chair in (9d).3
Therefore, a null comitative in a PC complement should be able to provide the adverb separately with the requisite plural NP. The result should be grammatical on the NP-dependent reading—but this prediction is disconfirmed. Note that (10) is grammatical without the adverb and that PRO is naturally interpreted as ‘Mary, the chair, and the dean’.4
(10) Mary told the chair and the dean that she prefers [PRO to meet (*separately) before Christmas].
We thus see that the null comitative analysis fails in three independent respects. PC complements contain neither a syntactically singular subject (which could bind a personal reflexive), nor a semantically singular subject (which could saturate a predicate of individuals), nor a silent comitative phrase (which could sanction the adverb separately). Rather, they behave exactly as expected if their subject is a genuine collective noun not associated with any comitative phrase.
2 NP Comitative Alternatives
While a null comitative phrase adjoined to VP cannot capture the facts of PC, perhaps a different implementation of the comitative analysis can. In particular, one could posit a comitative phrase adjoined to PRO itself, which would produce semantic plurality in PC. One such analysis was explicitly developed for PC by Rodrigues (2007) (see also Witkoś and Snarska 2009), and another one, which was developed for adnominal comitatives in Russian (McNally 1993, Dalrymple, Hayrapetian, and King 1998, Ionin and Matushansky 2003), can be straightforwardly extended to PC (as suggested by an anonymous reviewer).
Assuming the movement theory of control for PC, Rodrigues claims that PC emerges when a null pronoun adjoins to the embedded DP, creating a ‘‘big DP’’ with it: [DPpro DP]. The internal DP segment then raises to become the controller, stranding pro in the base thematic position (the embedded Spec,VP). Rodrigues likens this null pronoun to associative morphemes, found in languages like Japanese and Chinese; when attached to a name (say, John), these morphemes yield a plural denotation (in this case, ‘John and his associates’). Alternatively, a null adnominal comitative may adjoin to PRO as it does to overt NPs in Russian. The two analyses are depicted in (11).
Stranding a null associative morpheme
[TP Johni T [VP ti wants [TP ti to [VP [DPpro ti] meet]]]]
Generating a null adnominal comitative phrase
[TP Johni T [VP ti wants [CP [DP PROi [
with them]]j to [VP tj meet]]]]
Consider first analysis (11a). The singular number of PC PRO is explained by the fact that the embedded Spec,TP hosts a copy of a singular DP and not the complex big DP. In relation to the phenomena discussed above, the predictions of (11a) are not entirely clear, as it is not obvious which of the two subject positions—the embedded Spec,VP, which is plural, or the embedded Spec,TP, which is singular—is the criterial one. The null hypothesis should be that both positions are available and visible. If so, (11a) seriously overgenerates. A singular reflexive or secondary predicate should be able to take the singular subject in Spec,TP as antecedent—contrary to fact (see (6) and (8)). A syntactically plural big DP subject in Spec,VP should provide a distributive set for separately, also incorrectly (see (10)). In fact, as I show in Landau 2000, 2016, plural reflexives and secondary predicates are also ruled out in PC, given that PC PRO is syntactically singular. The big DP, however, being plural, should be able to license these elements from the embedded Spec,VP.5
*John prefers to introduce themselves/ourselves.
*John prefers to cooperate as partners.
The adnominal comitative analysis in (11b) is equally challenged by the absence of any evidence for syntactic plurality in PC complements, as overt NP comitatives license—in fact, require—plural agreement on verbs and reflexives/reciprocals (Dyła 1988, McNally 1993). Dyła and McNally have not discussed secondary predicates, but these forms too display plural agreement when predicated of [NP1with NP2] (Olga Kagan, pers. comm.).
Boris sfotografiroval Lenu s Katej golymi/*goloj.
Boris photographed Lena with Katja nude.PL/*F.SG
‘Boris photographed Lena and Katja nude.’
Furthermore, to the extent that the analyses in (11) are motivated by and therefore retain the properties of actual constructions with adnominal comitatives, they fail to capture a central semantic fact about PC, highlighted in Landau 2016: PC PRO denotes a group, not a sum, hence is not distributable. In contrast, associative plurals (strategy (11a)) clearly allow distributive readings in the presence of an overt distributor, as illustrated in (14) for Hungarian (Moravcsik 2003) and Japanese (Satoshi Tomioka, pers. comm.).
Distributive reading of associative plurals: Hungarian
Péter-ék különböző időben érkeztek.
Peter-ASS.PL different at.times arrived
‘Peter and his associates arrived at different times.’
Distributive reading of associative plurals: Japanese
Onnanoko-tati-wa sorezore tigau-iro-no syatu-o
girl-ASS.PL-TOP each differ-color-GEN shirt-ACC
‘The girls are each wearing shirts of different colors.’
Distributive reading of plural comitative: Russian
Petja s Vasej sideli na raznyx skamejkax.
Petja with Vasej sat on different benches
‘Petja and Vasja sat on different benches (from each other).’
(14)–(15) should be contrasted with (10) and (16a), where separately and different fail to distribute over PC PRO (different in (16a) only has a discourse-anaphoric reading). Note in particular that standard plural pronouns pattern with the comitative constructions in allowing distributivity (16b–c).
Johni wanted [PROi to (each) visit different cities].
≠ John wanted that each member of the group including him would visit a city different from the cities that other members visit.
John wanted them to (each) visit different cities.
John wanted them to (each) visit the city separately.
Evidently, then, the NP comitative analysis of PC fares no better than the VP comitative analysis. Even if one is willing to accept the idea that null comitative phrases are available in the grammars of many languages, such phrases should exhibit distributional and interpretive properties quite different from those observed in PC. Hence, these analyses do not contribute to our understanding of the nature of PC.
3 A Glimpse into Romance
One possible conclusion from the data considered in section 1 is that while the null VP comitative analysis is untenable for English, it is still a viable option for other languages. Indeed, this position is taken by Sheehan (2012, 2014). In particular, Sheehan claims that PC in Italian, French, Spanish, and (uninflected infinitives in) European Portuguese is exclusively formed with a null comitative phrase adjoined to VP. The main evidence for this claim is the selective availability of PC in these languages. According to Sheehan, PC is only attested with embedded predicates that can occur with a comitative phrase. For example, French se disputer ‘argue’ can take a comitative phrase, but se rencontrer ‘meet’ cannot, and therefore only the former may occur in a PC complement. English, as we saw in (3), does not respect this correlation.
While I cannot evaluate Sheehan’s evidence here, I would like to point out an immediate prediction her proposal makes: the interactions of the postulated null comitative structure that were attempted (and failed) in section 1 should succeed in the Romance languages. A semantically singular null subject and a syntactically present null pronominal in the comitative phrase should display the characteristic behavior of their overt counterparts.
As it turns out, at least French patterns with English in failing all the relevant tests that could diagnose a null comitative structure. As (17a) shows, the verb se réconcilier ‘reconcile’ is a comitative verb in French. Nevertheless, just like the subjects of its English counterparts in (6)–(7), when occurring inside a PC complement, its subject can neither bind a singular personal reflexive (17b) nor license a singular secondary predicate (17c); and just like its English counterpart in (10), it does not license the NP-dependent reading of séparément ‘separately’ in a PC complement (17d).
Jean s’est réconcilié *(avec Marie).
Jean SE-is reconciled with Marie
‘Jean reconciled *(with Marie).’
Jean a dit à Marie qu’il préférait ne pas
Jean has said to Marie that-he preferred NEG not
se réconcilier (*lui-même) ce soir.
to-reconcile himself this night
‘Jean said to Marie that he preferred not to reconcile (*himself ) tonight.’
Jean a dit à Marie qu’il était content
Jean has said to Marie that-he was happy
de se promener ensemble, enfin, (*en homme libre).
to-have-a-walk together finally (as man free)
‘Jean told Marie that he was happy to finally have a walk together (*as a free man).’
Marie a dit à son père et à sa mère qu’elle
Marie has said to her father and to her mother that-she
préférerait se réconcilier (*séparément) avant
would.prefer to-reconcile (separately) before
‘Marie told her father and her mother that she would prefer to reconcile (*separately) before Christmas.’
Exactly the same pattern is attested with other comitative verbs, like correspondre ‘correspond’.6
Such examples should be carefully tested in the other Romance languages for which the null comitative analysis has been proposed. Nevertheless, the French data already raise a very curious puzzle: why would the null comitative analysis give rise only to a PC reading and not to any other detectable syntactic or semantic effect? The ungrammatical versions of (17b–d) furnish a stronger negative argument against the null comitative analysis of French PC than the positive argument furnished by the alleged absence of PC with noncomitative verbs. The reason is that the latter fact does not exclude an alternative noncomitative analysis, whereas the former facts do exclude a comitative analysis (given the grammatical (5b), (7a), and (9a)).7 That being said, if PC in Romance is sensitive to the choice of embedded predicate in the way Sheehan suggests, this remains an open question for the noncomitative analysis.
Comitative constructions implicate a (possibly) singular subject and a comitative phrase. The comitative analysis of PC claims that both elements are present in PC complements, albeit in a null form. However, syntactic and semantic consequences of these elements when they are overt are systematically absent from PC, undermining the case for a common analysis. This is true both for English (pace Hornstein 2003 and Boeckx, Hornstein, and Nunes 2010) and for French (pace Sheehan 2012, 2014). The alternative analysis of PC posits no null comitative phrase and takes PRO to be a group-denoting, semantically plural noun. While this analysis should be further elaborated in view of the selective availability of PC in Romance, it is free of the problematic predictions that afflict the null comitative analysis.
I am grateful to Tal Siloni for comments and discussion on the topic of this squib.
1 Unlike what happens in split control, the additional controller in PC need not be a matrix argument or even linguistically present. Split control by two singular arguments triggers syntactic plurality in the complement (e.g., John proposed to Mary to become partners), but PC does not (see Landau 2000: 53–55, 2013:172–174). All the control predicates used in this squib are monotransitive; hence, none of them can be analyzed as cases of split control.
2 Comitative phrases, in fact, split into two types: the argumental one combines with inherently reciprocal verbs (like meet), and the adjunct one combines with any verb (like work). The two types differ in a number of additional respects (see Siloni 2008, 2012). However, since the null comitative analysis has been proposed for PC with both types of verbs, I henceforth disregard this distinction.
3 It is well-known that depending on lexical choices and dialects, collective nouns in English may, in fact, display syntactic plurality (see Levin 2001, Corbett 2006:211–213). The use of singular verbal inflection in (9b,d) is supposed to rule out this possibility. In general, to the extent that PC PRO is not syntactically plural in a given dialect, it is not expected to license the NP-dependent reading of separately in that dialect. Interesting implications arise for languages like Portuguese, where PC induces syntactic plurality in inflected infinitives (see Modesto 2010, Landau 2016, Sheehan to appear); I leave them for future work.
4 An anonymous reviewer remarks that in (10), a distributive reading is available to some speakers. The reason is plausibly related to the variation mentioned in footnote 3. If some speakers (e.g., of British English) generate PC PRO as a syntactically plural noun, a distributive reading will be licensed, just as it is licensed for these speakers with standard collective nouns, in the absence of singular inflection: for example, Her family (have) met separately.
6 The French data were tested with 5 native speakers; their judgments were completely uniform.
7 Of course, one might suggest that null comitative phrases are different from overt ones precisely in the empirical aspects investigated here. I take it that such a move would rob the analysis of its explanatory appeal.