This article investigates the validity of the theory of mediated predication by examining one of the proposed overt realizations of Pred0. Taking the law of parsimony as our starting position and using evidence from English, Russian, and Serbo-Croatian, we argue that the element that looks like the preposition ‘for’ is, in fact, a preposition (not Pred0), and we show how it explains the syntax and the semantics of the relevant ‘for’ sequences. Cases of apparently predicative interpretation of ‘for’-PP result from the interplay between the meaning of the preposition ‘for’ and the metaphorical reinterpretation of motion and locative verbs that ‘for’-PPs combine with.
On the basis of cases like (1a–c), Stowell (1981, 1983) argued that lexical projections other than VPs can have specifiers that function as their subjects. Stowell proposed that in examples like (1a–c) the main verb takes as its complement a bare lexical AP, NP, or PP, whose subject has moved to the matrix Spec,TP. The propositional meaning distinguishes small clauses from noun phrases, for which the existence of a specifier (hosting the possessor in the Anglo-Saxon genitive in English) was never in doubt (see Abney 1987). The term small clause was introduced to denote a constituent consisting solely of a subject and a (nonverbal) predicate, with no functional projections, such as tense, modality, or aspect.
Besides the classical instances of small clauses in the complement position of ECM (exceptional case-marking) and raising verbs, small clauses have been argued to appear in absolute constructions (2a), depictives (2b), expressives (2c), resultatives (2d), and some other environments (see Safir 1983, Simpson 1983, Moro 1995, Rothstein 2000, Potts and Roeper 2006).1
[With John sick], we’ll never get the job done on time.
John left the room [PRO angry].
[Me mad]?! Ridiculous!
They hammered [the metal flat].
In all these environments, the denotation of the postulated small clause is, as expected, propositional,2 and no functional material appears between the clearly identifiable subject and predicate. However, Bowers (1993, 2001) proposed that small clauses are in fact projections of a functional head, Pred0 (his Pr0), whose function is to turn its sister into a predicate. Hence, the predication is not direct (as under Stowell’s approach), but mediated by a functional head Pred0. It is the presence of a PredP that makes possible what would seem to be an instance of prohibited coordination of unlike categories (see also Sag et al. 1985, Bayer 1996).
I consider John crazy and a fool.
Bill is unhappy and in trouble.
Among arguments in favor of the existence of Pred0 one often finds data like (4a–b), where elements like for, as, into, and their crosslinguistic equivalents are taken to instantiate Pred0 overtly (se e, e.g., Emonds 1985, Bailyn and Rubin 1991, Aarts 1992, Bowers 1993, 2001, Bailyn 2001, 2002, Den Dikken 2006).
Mary takes him for a fool.
Jessamine views her mother as her best friend.
Sam considers [him a fool].
Jessamine made [her best friend jealous].
Case marking in Slavic languages has also been argued (Bailyn 1991, 2001, 2002, Bailyn and Rubin 1991, Bailyn and Citko 1999) to provide strong evidence in favor of analyzing small clauses as projections of a functional head: since in many Slavic languages the small clause predicate is marked instrumental, Pred0 has been argued to assign case to its sister. An identical structure is then postulated (Bailyn 2001) for the two examples in (7), which appear to have exactly the same meaning; Pred0 is null and case-assigning in (7a) but overt and case-absorbing in (7b).3
Taking the law of parsimony as our starting position and using evidence from Russian, Serbo-Croatian, and English, we will argue that the item that looks like the preposition za/for is, in fact, a preposition—not Pred0—and we will derive from this the putative case alternation in (7). We will also treat the semantic side of the issue, accounting for the derivation of the apparent propositional meaning of (7b), as well as for two additional environments, where za/for appears to introduce a small clause predicate. We will therefore show that the data at hand can be accounted for without the need to postulate a multiplicity of za/for-entities.
The article is organized as follows. In the remainder of this section, we provide a general introduction to the NP1-‘for’-NP2 construction in Russian, Serbo-Croatian, and English and list the theoretically possible structures for cases like (7b), discussing the implications these structures have with respect to the status of ‘for’, the interpretation of the entire sequence (NP1-‘for’-NP2), and the realizations available for NP1 and NP2. In section 2, we test the validity of these predictions for these structures. In section 3, we pursue the hypothesis that in examples like (4a) and (7b) ‘for’ is a preposition, which is not only the simplest hypothesis, as it does not stipulate a multiplicity of ‘for’-entities, but also the one that straightforwardly captures the relevant findings from section 2. Section 4 is the conclusion.
1.1 The Background of the NP1-‘for’-NP2 Construction
Examples like (8a–c) resemble examples like (9a–c), where an attitude verb has been hypothesized to take a small clause as its complement: both cases ascribe to the matrix subject an epistemic state where a predicational relation obtains between the accusative-marked NP following the verb and the remaining NP (or AP). As a result, it has been suggested that in examples like (8a–c) the main verb also combines with a small clause complement.
More specifically, the proposal that examples like (8a–c) involve small clauses has been suggested in order to support the hypothesis (Bowers 1993) that small clauses are necessarily headed by the functional head Pred0, serving as a mediator in the establishment of a predicational relation. In this context, the English for (Aarts 1992, Bowers 1993, Den Dikken 2006) and its crosslinguistic counterparts (e.g., Bailyn and Rubin 1991, Starke 1995, Bailyn 2002, Den Dikken 2006) have been offered as overt instances of the generally phonologically null Pred0.
To begin the discussion of the syntax of predicate-like ‘for’-PPs: initial evidence in favor of the small clause hypothesis comes from the fact that the property denoted by the complement of ‘for’ (NP2) can be predicated of the surface object or of the surface subject (NP1), thus paralleling ECM and raising small clause complements.
They took [NP1 Alexandra] for [NP2 a Spaniard].
[NP1 Alexandra] can pass for [NP2 a Spaniard].
The ability of the ‘‘direct object’’ NP1 to undergo passivization (11), pronominalization (12), and wh-movement (13) is compatible with its being either the direct object or the small clause subject. The fact that the putative predicate NP2 can be pronominalized (12a)/(14) and questioned (15) is less expected (see section 2.1).
His silence was taken/mistaken for consent.
first kiss SE take.3SG for anniversary.date.ACC
‘The first kiss is counted as the anniversary date.’
mistake.INSAUX.1SG exchange/take.PPRT.M.SG.PASS for spy.ACC
‘I was mistaken for a spy.’
spy AUX.PAST.SG accept.PASS.M.SG for self’s.ACC
‘The spy was taken for one of their own.’
The fact that NP2 is marked accusative in Russian and Serbo-Croatian does not, in and of itself, tell us much about its position in the structure (but see section 2.3). The same is true of the inability of the putative [NP1-‘for’-NP2] to undergo movement as a single constituent, since uncontroversial small clauses cannot be moved either. Both points are illustrated in (16) for English, but are valid also for Russian and Serbo-Croatian.
*[Alexandra for a genius] I took.
*[Alexandra a genius] I consider.
Finally, in English as well as Russian and Serbo-Croatian, the ‘for’-NP sequence yields a predicate-like interpretation only with a restricted set of verbs. Following the general trend, we will concentrate in this article on mental intake verbs (take, mistake, and pass in English; sojti ‘come off, pass’, prinjat’ ‘accept’, and sčitat’ ‘consider’ in Russian; and uzeti ‘take’, zameniti ‘exchange’, and proći ‘pass, go through’ in Serbo-Croatian), touching only lightly upon the two additional lexical-semantic groups of verbs that assign a predicate-like interpretation to za/for-PPs in some of the languages under consideration. One group, present only in Serbo-Croatian, consists of nomination verbs, like postaviti ‘appoint’, proglasiti ‘proclaim’, izabrati ‘choose/ elect’, and unaprediti ‘promote’. The other group is limited to the verbs have/imati, get/dobiti, and want/želeti in English and Serbo-Croatian. We discuss both groups in section 3.
Next, we present the various structures entertained for small clauses as a preliminary for a deeper investigation of ‘for’-PPs.
1.2 Potential Structures
The hypothesis that examples like (4a), (7b), and (8a–c) involve Bowers’s (1993) PredP entails that the English for and its crosslinguistic counterparts play the role of Pred0, functioning as a mediator in the establishment of a predicational relation.
Neither of the structures in (17) has been entertained for (8a–c). Stowell’s (1981, 1983) concept of small clauses as bare lexical NPs, APs, or PPs would place him/me/ego in (8) in the specifier of the relevant lexical head, therefore treating ‘for’ as a preposition, as in (17a). The symmetrical or exocentric structure in (17b), adopted by Moro (1997) (though for copular clauses only), also would not bestow any special status on ‘for’.
Conversely, the hypothesis that small clauses are projections of Pred0 makes available two potential sites for ‘for’. In the X-bar structure in (18), ‘for’ lexicalizes Pred0, whereas in the more articulated structure (19) (Starke 1995), it heads a CP-like functional projection above PredP.
The small clause analyses must be compared with two alternative analyses of the same examples: a ditransitive treatment of the verb, where the ‘for’-NP sequence corresponds to the PP argument of the verb, illustrated in (20a); and an adjunction treatment, illustrated in (20b), where the ‘for’-NP sequence, also forming a constituent, is a VP adjunct. Importantly, both analyses in (20) assume that ‘for’ performs its well-established and uncontroversial role as a preposition, which makes either of the two preferable to any analysis that needs to postulate multiple ‘for’-entities.
How do we distinguish among the six options available? As discussed above, both Stowell’s small clause and Bowers’s PredP have been introduced to account for syntactic constituents that can function as complements of ECM and raising verbs and that therefore have a proposition-like interpretation. Juxtaposing the small clause structures (17a–b), (18), and (19) with their two alternatives (20a–b), it is easy to see that the former make predictions that the latter do not, in particular regarding the status of NP2, the status of NP1, the status of ‘for’, and the interpretation of the NP1-‘for’-NP2 sequence.
Status of NP2: If ‘for’ is Pred0, its complement should be a predicate and cannot be a proper name, a (nonpredicative) pronoun or wh-word, or a quantified NP.
Status of NP1: If the NP1-‘for’-NP2 string is a small clause, NP1 is its subject. If we are dealing with a single predicate, then NP1 is an object of the verb.
Status of ‘for’: Under the analyses in (18) and (19), ‘for’ is either a Pred0 or some sort of C0. Under the bare small clause analyses in (17a–b), as well as under the single-predicate analyses in (20a–b), it is a preposition and therefore expected to combine with NPs only and to assign case.
Semantics of the NP1-‘for’-NP2 sequence: Under the small clause analyses, the main verb combines with a proposition (or a similar semantic entity; see footnote 2). This predicts that (a) verbs that appear with the NP1-‘for’-NP2 sequence should also take regular small clauses and perhaps also TPs and/or CPs; (b) conversely, verbs that normally take small clauses (change-of-state or intensional verbs) should combine with the NP1-‘for’-NP2 sequence.
In section 2, we will show that none of the predictions of the small clause analyses holds for mental intake verbs in English, Russian, or Serbo-Croatian: ‘for’ behaves like a preposition and the NP1-‘for’-NP2 string does not behave like a small clause. In section 3, we provide our own analysis, arguing for structure (20a) for mental intake verbs (section 3.2), as well as for other verb-‘for’ combinations (section 3.3 and 3.4).
2 Building a Case for a Preposition
2.1 Lexical Restrictions on ‘for’-PPs
Given the wide range of environments where small clauses can appear, it is natural to ask first whether the putative Pred0 ‘for’ appears in canonical small clause environments. As discussed above, the ‘for’-NP sequence yields a predicate-like interpretation only with a restricted set of verbs. In other small clause environments ‘for’-PPs are impossible, as illustrated in (21)–(23) with attitude verbs, depictives (for Russian and English only), and change-of-state verbs.4
Besides being excluded in small clause environments, za/for-PPs are optional: as the glosses above suggest, the verbs compatible with the za/for-PP can also appear without it. The result is more or less felicitous depending on the lexical choice for the direct object (or, in the case of pass, the subject): for instance, silence cannot be taken or mistaken (24a), while roads and entrances can be (24b). As shown in (25), in the absence of a for-PP (mis)take and pass can no longer function as attitude verbs and, being eventive, are ungrammatical in the present tense in English (generic interpretation is excluded pragmatically here).5 As the glosses above suggest, the verbs compatible with the za/for-PP can also appear without it ((26)–(27)), but then they no longer function as attitude verbs ((24), (26)–(27)).
Jeremy took/mistook their silence *(for consent).
I mistook the road/the entrance.
*Roger mistakes the entrance.
*Juliette takes the right road.
#Ursula passes (by the cathedral).
This lexical restriction on the use of ‘for’-PPs constitutes our first piece of evidence against the hypothesis that ‘for’ in English, Serbo-Croatian, and Russian in examples like (8a–c) and (21a–c) is the head of the small clause, Pred0. Indeed, a functional projection with a particular realization of its head is not expected to appear with a handful of verbs only. This conclusion is reinforced by the fact that only one of these verbs, the Russian sčitat’ ‘count, consider’, is also compatible with small clause and CP complements; we return to this issue in section 2.6.
2.2 Pred0 as a Mediator and the Semantic Type of NP2
By definition, canonical small clauses consist of a subject, which may be pronominal or quantified, and a predicate; see (28). If the predicate is an NP, it cannot be replaced by a pronoun, a proper name, or a quantified NP, as none of these are semantically predicates.6
I consider [John a fool/*each student/*him].
I was made [president/an actor/*every student/*him].
This restriction clearly does not hold for NP1-‘for’-NP2 sequences, since NP2 can easily be a pronoun, a proper name, or a quantified NP, as illustrated in (29) and (30). (English test cases are provided by the translations of the Russian and Serbo-Croatian examples.)7
Since in the bare small clause structures in (17a–b) the entire ‘for’-PP rather than NP2 is taken to be the predicate of the small clause, the data in (29) and (30) are irrelevant for these analyses. In the CP structure in (19), on the other hand, ‘for’ is taken to be external to the small clause and the contrast is inexplicable. Finally, for the X-bar structure in (18) a possible way out may lie in hypothesizing that the semantic import of the covert Pred0 and the overt ‘for’ is not the same. Taking this route is complicated by the fact that different authors do not appear to make the same assumptions about the semantic role of Pred0. While Bowers (1993, 2001) suggests that Pred0 combines with its complement to return a predicate (semantic type ‹e, t›), Bailyn (e.g., 2001, 2002) and Den Dikken (2006) seem to consider its function purely syntactic: the sister of Pred0 is taken to be a predicate. Examples like (29a–b) and (30a–b) show that the latter position cannot be extended to the alleged ‘for’-PredPs. The former position (i.e., that different lexicalizations of Pred0 do not play the same semantic role) simply restates the dilemma: any preposition or transitive verb can combine with a referential or quantified NP and return a semantic predicate, which does not warrant calling it Pred0—why should ‘for’ be any different?
Another piece of evidence favoring the referential status of NP2 is that the wh-word used to replace it in (13a–c) is who, rather than what. As is well-known (see, e.g., Williams 1983), the choice of the wh-word differentiates between the predicative and argumental/referential uses of [+human] NPs, with what, instead of who, used for [+human] NP predicates. (31) and (32) show clear-cut predicational cases in English.
(31) What/*Who is she by profession?
What did John become? A doctor.
?What did John talk to? A doctor.
(adapted from Williams 1983:(11a–b))
The same is true in Serbo-Croatian for the feature [+human]: the use of ko ‘who’ in (33) and
If the wh-word for properties related to [+human] entities is ‘what’, rather than ‘who’, then fact that ‘who’ is used in (15) indicates that the sister of ‘for’ is a referential rather than predicative NP, on par with the uncontroversially referential NP1 in (13).8 In other words, the choice of the wh-word in English and Serbo-Croatian also argues against treating NP2 as a predicate.
2.3 The Status of NP1
In this section, we test how NP1 behaves with respect to two diagnostics that distinguish subjects from direct objects: subextraction and anaphor binding. After demonstrating that for principled reasons, the availability of subextraction out of NP1 does not allow us to conclude anything, we show that NP1 does not behave like a subject with respect to anaphor binding.
2.3.1 The Subject Condition
Our first argument against the small clause analysis of ‘for’-PPs comes from subextraction. As Kayne (1984) notes, postverbal NPs with consider-type verbs behave like subjects of full and infinitival clauses in disallowing subextraction (Chomsky’s (1973) Subject Condition), which has been taken to mean that such NPs are genuine subjects of small clauses.
*Whom does Mary consider [SC[NP friends of *t] idiotic]?
*Of whom does Mary consider [SC[NP friends *t] idiotic]?
If NP1-for-NP2 sequences are small clauses, then NP1 should be an island for extraction. The facts do not confirm this, however, since extraction out of NP1 leads to a sentence that is degraded, but still grammatical.
??Whoi does Mary take [friends of ti] for fools?
?Whoi did Clarissa mistake [a portrait of ti] for a picture of Cher?
Why do NP1-for-NP2 sequences behave differently from small clauses? Following one line of reasoning, the contrast could be due to the choice of the verb. As Basilico (2003) notes, extraction out of small clause subjects of perception and causative verbs is much better than extraction out of small clause subjects of opinion verbs.
??Which subjecti do you consider [a book about ti] too boring for your class?
??Whoi did you find [a photograph of ti] rather unattractive?
??Whoi did you judge [a rumor about ti] false?
(Basilico 2003:5, (11))
Which planeti did you see [a picture of ti] appear on your computer screen?
Whoi did you let [a rumor about ti] spread around the entire department?
Which presidenti did you watch [a picture of ti] burn in the wastebasket?
(Basilico 2003:5, (12))
We conclude that subextraction facts neither disconfirm nor support the hypothesis that NP1 is a small clause subject. Next, we turn to anaphor binding, which also distinguishes subjects from direct objects.
2.3.2 Anaphor Binding in English
Our second argument against the small clause analysis of ‘for’-PPs comes from anaphor binding. It has long been noted (see Stowell 1981, 1991) that subjects of small clauses can induce opacity effects for the binding of anaphors and pronouns. In English, the subject of a small clause must bind anaphors and cannot bind pronouns in the small clause predicate, while true direct objects do not preclude binding by the subject.
Clarissai considers Virginiaj kind to herself*i/j.
Clarissai considers Virginiaj too kind to heri/*j.
Clarissai introduced Virginiaj to herselfi/j.
At first blush, English anaphor binding might seem to provide clear evidence against analyzing the NP1-for-NP2 sequence as a small clause, since NP1 does not give rise to opacity effects.9
(40) Clarissai took/mistook Virginiaj for a big fan of herselfi/j.
The situation is more complicated, however, since examples (39a–c) and (40) do not form a minimal pair: the anaphors in the former form part of AP predicates, while the latter has an NP-internal anaphor. Strikingly, in small clauses with NP predicates, the subject ceases to be opaque and the contrast among uncontroversial small clauses (41a), true ditransitive verbs (41b), and take/mistake cases (40) disappears.
Clarissai considers Virginiaj a big fan of herselfi/j.
Clarissai showed Virginiaj to herselfi/j in the mirror.
We conclude therefore that English anaphor binding does not appear to distinguish the small clause subject and the direct object. The situation differs in Slavic languages, where the subject orientation of reflexives is absolute; that is, only a local subject can serve as a binder for a reflexive (e.g., Rappaport 1986, Reinders-Machowska 1991, Progovac 1992, 1993, Zlatic 1996). In other words, in Slavic the direct object cannot bind anaphors that it c-commands while the small clause subject can. As a result, we can diagnose small clause subjects by the binding possibilities they introduce. Since Russian and Serbo-Croatian differ with respect to certain details, we examine them separately.
2.3.3 Anaphor Binding in Serbo-Croatian
In Serbo-Croatian, the φ-deficient reflexive pronoun sebe ‘self ’ is used for all genders, persons, and numbers. As (42a–b) show, as in English, the reflexive must be bound within its binding domain and, unlike in English, it must be bound by the subject; no other c-commanding NP counts. Furthermore, the subject need not necessarily be in the nominative case, since a quirky-case subject, dative (43a) or accusative (43b), can also act as a binder. The contrast between the dative complement in (42b) and the dative subject in (43a) is very striking in this respect.
Tristrami tvrdi da je Klarisaj videla sebe*i/j/k.
Tristram claim.3SG that AUX.3SG Clarissa see.PPRT.F.SG self.ACC
‘Tristram claims that Clarissa saw herself.’
Tristrami je Klarisij pokazao sebei/*j/*k u ogledalu.
Tristram AUX.3SG Clarissa.DAT show.PPRT.M.SG self.ACC in mirror
‘Tristram showed Clarissa himself in the mirror.’
Lorensu/Svakome je bilo žao sebe.
Lawrence.DAT/everybody.DATAUX.3SG been.N.SG sorry self.GEN
‘Lawrence/Everybody felt sorry for himself.’
Tristrama/Svakog je strah od sebe.
Tristram.ACC/everybody.ACCAUX.3SG fear from self.GEN
‘Tristram/Everybody fears himself.’
The possessive pronoun svoj ‘self ’s’ behaves like sebe ‘self ’ in its strict subject orientation, as shown by the ungrammaticality of (44), where the inanimate subject is excluded as a possible binder on pragmatic grounds.
*Taj dogadjaj je pokazao Lorensu svog najvećeg neprijatelja.
that incident AUX.3SG show.PPRT.M.SG Lawrence.DAT self ’s.ACC biggest.ACC enemy.ACC
Intended: ‘That incident showed to Lawrence his biggest enemy.’
Just like the subjects of the ‘that’-clauses in (42a) and (45a) and unlike the indirect object in (42b), the small clause subjects in (45b) and (45c) are legitimate binders for a reflexive in an AP or NP predicate in Serbo-Croatian.10
Tristrami smatra da je Klarisaj ljuta na sebe*i/j.
Tristram consider.3SG that AUX.3SG Clarissa.NOM angry on self
‘Tristram thinks that Clarissa is angry at herself.’
Tristrami smatra Klarisuj ljutom na sebe*i/j.
Tristram consider.3SG Clarissa.ACC angry on self
‘Tristram considers Clarissa angry at herself.’
Tristrami čini Klarisuj nezadovoljnom sobom*i/j.
Tristram make.3SG Clarissa.ACC dissatisfied self.INS
‘Tristram makes Clarissa dissatisfied with herself.’
Taj dogadjaj je učinio Lorensa svojim najvećim neprijateljem.
that incident AUX.3SG make.PPRT.M.SG Lawrence.ACC self ’s biggest enemy
‘That incident made Lawrence his own worst enemy.’
Having established that Serbo-Croatian small clause subjects can bind reflexives in the predicate, we can now test whether the NP1-za ‘for’-NP2 sequence is a small clause: can NP1 bind an anaphor in NP2? It turns out that although NP2 can contain an anaphor or be an anaphor itself, that anaphor cannot be bound by NP1.
Tristrami je greškom zamenio Klarisuj za sebei/*j.
Tristram AUX.3SG mistake.INS exchange/take.PPRT.M.SG Clarissa.ACC for self
‘Tristram mistook Clarissa for himself.’
Tristrami je greškom zamenio Klarisuj za svogi/*j neprijatelja.
Tristram AUX.3SG mistake.INS exchange/take.PPRT.M.SG Clarissa.ACC for self ’s.ACC enemy
‘Tristram mistook Clarissa for his enemy.’
2.3.4 Anaphor Binding in Russian
The Russian nominal reflexive sebja ‘self ’ and possessive reflexive svoj ‘self ’s’ are also subject-oriented, and the binding domain is delimited by a finite subject.
Milicionerj rassprašival arestovannogoi o sebe*i/j.
policeman.NOM question.PAST.M.SG arrested.ACC about self.LOC
‘The policeman questioned the suspect about himself.’
My dovezli rebënka do svoego doma.
we.NOM drive.PAST.PL child.ACC until self ’s home.GEN
‘We drove the child to our/*his home.’
(Rappaport 1986:101, (7) and (8); glosses and translations slightly adjusted)
Generali ne razrešaet sekretaršej [PROj pozvolit’ dvornikuk [PROk nazyvat’ sebjai,j,k,*l Valej]].
general NEG allow.3SG secretary.F.ACC permit.INF yard.keeper.DAT call.INF self.ACC Valya.INS
‘The general does not allow the secretary to permit the yard-keeper to call him/her/himself Valya.’
Anaphor binding by the small clause subject is not accepted in equal measure by all speakers of Russian if the anaphor is the simple reflexive sebja ‘self ’ or svoj ‘self ’s’. However, reinforcing the reflexive with an intensifier makes such binding possible for all speakers (see Lyutikova 1998).
Sonjaj sčitala egok ravnodušnym . . .
Sonya consider.PAST.F.SG him.ACC indifferent
‘Sonya considered him indifferent toward herself.’
daže k sebe*j/%k.
even toward self.DAT
‘Sonya considered him indifferent even toward himself/*herself.’
daže k %samoj/√samomu sebe.
even toward EMPH.F/M self
‘Sonya considered him indifferent even toward himself.’
k svoimj/%k detjam.
toward self ’s children
‘Sonya considered him indifferent even toward his/her children.’
daže k svoim*j/k (sobstvennym) detjam.
even toward self ’s own children
‘Sonya considered him indifferent even toward his own children.’
Sonjaj sčitala egok ugrozoj dlja (%samoj/samogo) sebjaj/%k.
Sonya consider.PAST.F.SG him.ACC threat for EMPH.F/M self
‘Sonya considered him a threat to herself/himself.’
Neither the simplex nor the reinforced reflexive can be bound by NP1 in the NP1-za ‘for’- NP2 sequence, showing that NP1 is as unlikely to be a small clause subject in Russian as it is in Serbo-Croatian.
Lolaj ne prinjala Markak za ugrozu dlja svoixj/*k detej.
Lola NEG accepted Mark for threat for self ’s children
‘Lola didn’t take Mark for a threat to her/*his children.’
Lolaj ne prinjala Markak za ugrozu dlja (samoj/*samogo) sebjaj/*k.
Lola NEG accepted Mark for threat for EMPH.F/M self
‘Lola didn’t take Mark for a threat to herself/*himself.’
We conclude that anaphor binding in Russian, like anaphor binding in Serbo-Croatian, shows that there is no small clause involved in the NP1-za/for-NP2 sequence. In the absence of a small clause, za/for is highly unlikely to be a Pred0.
2.4 The Status of ‘For’
The c-selectional properties of ‘for’ strongly suggest that it is not Pred0. As (52) shows, regular small clauses can have predicates of any lexical category, although some verbs appear to constrain their small clause predicates (Stowell 1981, 1983).
The hypothesis that za/for is a regular preposition in (52) and (53) is confirmed by the case it assigns: in its locative use the preposition za ‘for’ assigns instrumental case in both Russian and Serbo-Croatian, while in its other uses it assigns accusative.12 Given that the case on predicative NPs and APs in Russian and (to a lesser extent) in Serbo-Croatian is instrumental,13 accusative case marking is not expected if za ‘for’ is a Pred0.14
Finally, as Starke (1995) notes, ‘for’ in the NP1-‘for’-NP2 construction patterns with the preposition ‘for’ with respect to stranding: notably for English, while the preposition for can be stranded, the complementizer for cannot be (although, to be sure, this could be because the NP linearly following for is an embedded subject and extraction would violate the that-trace condition).
Who do you take me for?
*Who is it the time for _____ to leave?
The c-selectional properties of the ‘‘predicational’’ za/for, its phonological identity to the directional preposition za/for, and its ability to be stranded and to assign accusative case all point to its prepositional status, without excluding the bare small clause analyses in (17), where ‘for’ is assumed to be a preposition anyway. However, if ‘for’ lexicalizes Pred0, as in (18), or if it is a complementizer, as in Starke’s (1995) structure (19), at least some of its similarities to the preposition ‘for’ require an independent explanation.
Under the assumption that maximal projections can be coordinated only if they belong to the same lexical category, the postulation of PredP (see (56b)) explains why predicates can be coordinated (Bowers 1993).15
Supposedly impossible: Coordination of unlikes
*[AP crazy] and [NP a fool]
OK: Coordination of likes with PredP
[PredP [AP crazy]] and [PredP [NP a fool]]
However, if the sequence NP1-‘for’-NP2 is a small clause, coordinating it with a regular small clause should be possible. While in English no verb can both combine with a small clause and appear with an NP1-‘for’-NP2 construction, such is not the case in Russian and Serbo-Croatian. As the following examples show, a number of Serbo-Croatian verbs combine with both regular small clauses and NP1-za ‘for’-NP2 sequences (see section 3), but do not allow their coordination. The same is true for the only Russian verb compatible with both NP1-za ‘for’-NP2 sequences and regular small clauses, sčitat’ ‘count, consider’ (cf. (7)).
proclaim.PPRT.M.PLAUX.3PL him.ACC king.INS/for king.ACC
‘They proclaimed him king.’
*Proglasili su ga za kralja i vlasnikom zemlje.
proclaim.PPRT.M.PLAUX.3PL him.ACC for king.ACC and owner.INS land.GEN
Intended: ‘They proclaimed him king and the owner of the land.’
*Proglasili su ga vlasnikom zemlje i za kralja.
proclaim.PPRT.M.PLAUX.3PL him.ACC owner.INS land.GEN and for king.ACC
Intended: ‘They proclaimed him the owner of the land and king.’
consider.1SG Mary.ACC nice.INS/for pretty.ACC woman.ACC
‘I consider Mary nice/a pretty woman.’
*Smatram Mariju za lepu ženu i prijatnom.
consider.1SG Mary.ACC for pretty.ACC woman.ACC and nice.INS
Intended: ‘I consider Mary a pretty woman and nice.’
*Smatram Mariju prijatnom i za lepu ženu.
consider.1SG Mary.ACC nice.INS and for pretty.ACC woman.ACC
Intended: ‘I consider Mary nice and a pretty woman.’
we consider.1PL Mary.ACC beauty.INS/for smart.one.ACC
‘We consider Mary a beauty/a smart person.’
*My sčitaem Mariju krasavicej i za umnicu.
we consider.1PL Mary.ACC beauty.INS and for smart.one.ACC
Intended: ‘We consider Mary a beauty and a smart person.’
*My sčitaem Mariju za umnicu i krasavicej.
we consider.1PL Mary.ACC for smart.one.ACC and beauty.INS
Intended: ‘We consider Mary a smart person and a beauty.’
Starke’s (1995) structure in (19) provides the only small clause analysis that can deal with both sets of coordination facts on Bowers’s (1993) assumptions, since it assumes that all small clauses are CPs. In this case, the coordination of unlike categories is not at issue because the impossibility of coordinating regular small clauses and NP1-‘for’-NP2 sequences can be derived by assuming that the complementizers in question do not have the same featural specifications. As (60a–c) show, CPs headed by finite and nonfinite complementizers cannot be coordinated.
It is good for him to leave.
It is good that she is happy.
*It is good for him to leave and that she is happy.
The problem, however, is that Slavic languages provide no evidence for treating za ‘for’ as a complementizer, since it never appears in structures like (60a). Furthermore, no features have been independently suggested for the covert Pred0 and the overt ‘for’ to differ in.
The four structures discussed in section 1.2 all predict that the NP1-‘for’-NP2 sequence, being a small clause, should be propositional, but they differ on how this result is obtained. Postponing the latter issue until section 3, we can easily demonstrate that NP1-‘for’-NP2 sequences do not behave as if they were propositions. The verbs that such sequences appear with (‘take’, ‘pass’, ‘mistake’, etc.) are incompatible with regular small clauses, or indeed with any propositional complements (61); augmenting the alleged small clause with be, as in (62), is likewise disallowed.
*We took [him smart/a fool/in love].
*Hei can pass [ti smart/a fool/in love].
*We mistook [that the matter was serious].
*Watson mistook Sherlock Holmes for to be/to for be/to be for a criminal.
Beyond disallowing ECM/raising infinitives, Russian and Serbo-Croatian do not differ from English in this respect.
Uzimamo naš prvi poljubac za godišnjicu/*godišnjicom. Serbo-Croatian
take.3PL our.ACC first.ACC kiss.ACC for anniversary.ACC/INS
‘We take our first kiss as/for our anniversary.’
Conversely, verbs that normally take small clauses (attitude verbs or change-of-state verbs) systematically do not appear with ‘for’-PPs (with the exception of the Russian verb sčitat’ ‘consider’ (see footnote 3) and the Serbo-Croatian verbs discussed in section 3).
Jane made/considered him (*for) a fool.
He seems/became (*for) a fool.
he seem.PAST.M.SG/turn.out.PAST.M.SG fool.INS
‘He seemed/turned out to be a fool.’
*On kazalsja/okazalsja za duraka.
he seem.PAST.M.SG/turn.out.PAST.M.SG for fool.ACC
My našli ego zabavnym/??durakom.
we find.PAST.PL him.ACC amusing.INS/fool.INS
‘We found him amusing/a fool.’
*My našli ego za zabavnogo/duraka.
we find.PAST.PL him-ACC for amusing.ACC/fool.ACC
The observed double dissociation between classical complement small clauses and the alleged ‘for’-PredPs is not predicted by any of the small clause analyses discussed in section 1.2. Though it could be stipulated that the verbs appearing with NP1-‘for’-NP2 sequences l-select ‘for’ as the only possible head of their complement, the inability of the alleged ‘for’-PredPs to appear with attitude verbs remains a mystery. The inability of NP1-‘for’-NP2 sequences to function as adjunct small clauses of any kind is also surprising. In the next section, we show that the assumption that ‘for’ is a regular preposition and the ‘for’-phrase is a PP argument or adjunct can account for all the facts discussed above.
3 Deriving the Properties of ‘for’-PPs
Among the many interpretations of the polysemous preposition ‘for’ are purpose (68a), goal (68b), and substitution (68c). In this section, we will argue that these interpretations are sufficient to explain the predicate-like interpretation of ‘for’-PPs in English, Russian, and Serbo-Croatian.
What is necessary for this result?
Nicholas has left for Paris.
I’ll finish the job for you.
If we assume that ‘for’ is a preposition, there is no need to explain why a Pred0 (or a C0) should look like a preposition. The c-selectional and case-assigning properties of za/for also follow. Finally and most importantly, as we will argue in sections 3.1 and 3.2, it is possible to construct the mental intake interpretation for ‘(mis)take’ and ‘pass’ on the basis of the lexical semantics of these verbs in Russian and in Serbo-Croatian and the preposition za ‘for’ in its directional use, as in (68), where it assigns accusative case. We will extend our analysis to the two other groups of verbs that induce a predicate-like interpretation of za/for-PPs (sections 3.3 and 3.4).
3.1 Mental Intake Verbs
To derive the apparently predicative interpretation of ‘for’-PPs, we rely on the interaction of the meaning of the preposition ‘for’ with the metaphorical reinterpretation of the verbs ‘pass’ and ‘take’ as predicates of metaphorical rather than physical change of location.16 As noted above and as shown in (69)–(70), one of the basic meanings of the preposition ‘for’ is ‘in exchange for, instead of, in place of ’, though its distribution differs somewhat in the three languages.
Moreover, ‘take’ (though not other verbs interpreted intensionally with ‘for’-PPs) can be interpreted as an attitude verb in the absence of the ‘for’-PP.17
We took him to be a fool/sincere.
We took it as understood that you would come with us.
One more step of reasoning is required here: if an entity x is perceived in place of or in exchange for an entity y, then the subject of perception will, under normal circumstances, conclude that he or she is perceiving y. This is precisely how ‘take for’ is interpreted.
We hypothesize that it is this reasoning that facilitates the interpretation of the English pass, the Russian sojti ‘come off, pass’ and ščitat’ ‘consider, count’, and the Serbo-Croatian zameniti ‘exchange’ and proći ‘go through, pass’ as mental intake verbs in the environment of a ‘for’-PP. The question now arises how the ‘for’-PP integrates into the argument structure of the reinterpreted verb.
3.2 Argument or Adjunct?
In this section, we will argue that the ‘for’-PP must be viewed as an argument of its verb and that therefore, the structure in (20a) must be chosen over the structure in (20b). Our primary motivation for this choice comes from the fact that an NP complement of ‘for’ need not assert the existence of an individual of which the property denoted by the NP holds, and a definite NP complement of ‘for’ need not presuppose it.
As (73a–c) show, passing for an alien does not entail the existence of one, and somebody’s being taken for your sister does not entail that you have a female sibling. The nonexistential interpretation of an indefinite NP2 might, therefore, have been taken as evidence that it denotes a predicate, though to the best of our knowledge no such argument has been made. To obtain the same result without appealing to the small clause hypothesis, we need to assume that the verb takes scope over the ‘for’-PP, which is only possible if the ‘for’-PP is an argument rather than an adjunct. Independent evidence for the availability of such argument structure comes from the fact that in their literal interpretation the verbs ‘take’ and ‘pass’ can (sometimes must) also take a goal argument.
We took him home/to the park.
He passed into legend/into the next room/away.
Crucially for our proposal, while the language-specific choice of verbs combining with ‘for’- PPs and the consequent change in their interpretation are taken here to be idiosyncratic and not syntactically encoded, we emphasize that we view the syntax and semantics of the preposition ‘for’ as completely regular. By assuming that the ‘for’-PP is merged as the goal argument of what is essentially a motion verb, as in (20a), we can simultaneously account for its role in the reinterpretation of ‘(mis)take’ and ‘pass’ as mental intake verbs and for the predicate-like interpretation of the complement of ‘for’.
Next, we will provide additional evidence for the consistently prepositional status of za ‘for’ by discussing another environment in Serbo-Croatian (not available in either English or Russian) where the directional za ‘for’-PP is productively used to yield an apparently predicative meaning.
3.3 Nomination Verbs and the Goal Interpretation
As the following Serbo-Croatian examples show, with certain nomination verbs (such as proglasiti ‘proclaim’, postaviti ‘appoint’, krunisati ‘crown’, and izabrati ‘choose/elect’) the NP denoting the function/position assumed may (as in Russian) appear in the instrumental case or (unlike in Russian) be introduced by the preposition za ‘for’:
To explain this difference between Russian and Serbo-Croatian, we appeal to the fact that the directional use of za ‘for’ differs in the two languages. While in Russian the use of za ‘for’ is restricted to ‘behind, beyond’ (78a), in Serbo-Croatian its lexical meaning, at least in the directional use, is less specific. With Serbo-Croatian equivalents of motion verbs such as move, go, transfer, run, walk, swim, ride, drive, fly, and travel, it suggests movement toward a specific point (78b).
We hypothesize therefore that Serbo-Croatian nomination verbs are compatible with two argument structures, yielding the two different options for NP2. The instrumental case marking on NP2 indicates that the nomination verb takes a small clause complement, while the presence of za ‘for’ diagnoses a ditransitive structure. While in the former case the nomination verb is molded in the pattern of make, in the latter it is assimilated to put. As corroboration for this claim, note that in the environment of a za ‘for’-PP, the verb postaviti ‘put, set’, otherwise a typical locative verb (79a), is reinterpreted as ‘elect, appoint’ (79b). Thus, it undergoes a concrete-toabstract meaning shift: instead of describing a caused displacement of a concrete object to another physical location, it retains the same argument structure with the metaphorical endpoint of a ( preexisting) job position or function. Again, an apparently predicative interpretation arises from the interaction of the meaning of the preposition za ‘for’ and a metaphorical reinterpretation of the verb. The fact that the verb postaviti ‘put, set’ cannot take an NP marked with the predicative instrumental case further supports the insight that the interpretation of the za/for-NP is not predicative.
Further evidence for the directional interpretation of the za ‘for’-PP in (77) comes from the facts that (i) with the nomination verb unaprediti ‘promote’, the preposition za ‘for’ alternates with the preposition u ‘in(to)’ (80), and (ii) in examples like (81) the za ‘for’-NP is clearly nonpredicative as a result of both being universally quantified and not denoting a set of entities with the function in question.
Nominovali su ga za sve počasne funkcije u okviru Akademije.
nominate.PPRT.M.PLAUX.3PL him.ACC for all honorary.ACC functions.ACC within Akademija.ACC
‘He was nominated for all honorary functions in the Akademija.’
To conclude, treating ‘for’ as a preposition rather than Pred0 or C0 allows us not only to straightforwardly account for the cases where NP1-‘for’-NP2 sequences have an apparently propositional meaning in English and in Russian, but also to explain why Serbo-Croatian has an additional environment where this effect obtains. We hypothesize that generalizing from such change-of-state cases may lead to the reanalysis of a preposition as a copular particle—a stage that neither Russian nor Serbo-Croatian has yet achieved.
3.4 Wanting, Getting, and Having It All
The Serbo-Croatian examples and their translations in (82) illustrate yet another previously undiscussed environment in Serbo-Croatian and English (but not in Russian) where a za/for-PP gives rise to a predicate-like interpretation.
With all three verbs, NP2 must be indefinite even when denoting a singleton set (e.g., husband),18 and it should be interpretable relationally. The latter fact, along with a number of others, leads to the conclusion that (82a–c) all share the same structure, with the possession relation, as in (82a), serving as its core (see, e.g., McCawley 1974, Ross 1976, Dowty 1979, Den Dikken, Larson, and Ludlow 1996, Fodor and Lepore 1998, Harley 2004). We further hypothesize that the NP complement of za/for is interpreted here as denoting a function or role rather than an entity (or set of entities) and that the preposition za/for introduces a purpose. As the Serbo- Croatian examples and their English translations in (83) show, NPs denoting positions or roles are possible complements of za/for under this interpretation; the identity copular clause in (84a) and the appositive oblique in (84b) provide further evidence for the hypothesis that relational nouns of this type may also denote functions or roles.
As a result, we account for the indefinite article characterizing the NP complements of ‘for’ with ‘have’, ‘get’, and ‘want’ in English: as (84a–b) show, even NPs corresponding to functions appear with the indefinite article in the absence of an internal argument.
The final question to resolve is what configuration corresponds to the NP1-za ‘for’-NP2 sequence with imati ‘have’ in (82a) and, therefore, with dobiti ‘get’ and želeti ‘want’ in (82b) and (82c), respectively. If the za/for-PP indeed functions as a purpose specification, a small clause structure seems unlikely, even though have can in principle combine with PP small clauses (Harley 1998, Sæbø 2009).
We have our man in Havana.
They have a first-class specialist in Rosalie.
However, even if our semantic treatment is incorrect, za/for cannot be argued to function as Pred0 in examples like (82a–c). Indeed, in English, have can only combine with prepositional small clauses, and the hypothesis that za ‘for’ is Pred0 in (82a–c) violates this generalization.
(86) *We have our man tired/sick/happy/a fool/an engineer.
We conclude that this lexical-semantic verbal class also provides no evidence for treating ‘for’ as Pred0; rather, it argues for the structure in (20a).
In this article, we have argued that the hypothesis that ‘for’ heads a small clause has nothing to recommend it and much to disprove it. We have offered an alternative—the parsimonious view of ‘for’ as a preposition—and shown how it explains both the syntax of the NP1-‘for’-NP2 sequence and its semantics. We have argued that the ‘for’-PP can be merged as the goal argument of what is essentially a motion verb. This allows us to simultaneously account for its role in the reinterpretation of ‘(mis)take’ and ‘pass’ as mental intake verbs and for the predicate-like interpretation of the complement of ‘for’. Since under our account NP1-‘for’-NP2 sequences do not have a propositional meaning, their incompatibility with the verbs that take classical small clause complements follows straightforwardly. The hypothesis that za ‘for’ is a preposition also makes it possible for us to explain why Serbo-Croatian, unlike Russian or English, allows NP1-za ‘for’-NP2 sequences with nomination verbs by linking the hypothesized modification of the argument structure of such verbs to the independently attested goal interpretation of za ‘for’ in Serbo-Croatian. Finally, our proposal has the additional advantage of being able to deal with two previously unnoticed cases of predicate-like interpretation associated with ‘for’-PPs: with nomination verbs in Serbo-Croatian and with imati/have, dobiti/get, and želeti/want in Serbo- Croatian and English.
For comments and much pertinent discussion, we are grateful to Eddy Ruys, Joost Zwarts, and the audiences at the workshop ‘‘Syntax and Ontology of Predication’’ ( Paris, 7 February 2009) and the GLOW 33 workshop ‘‘Slavic Syntax and Semantics’’ (Wroclaw, 13 April 2010). We also thank one of the two LI reviewers for a wealth of extremely useful and insightful comments. Both authors’ work was supported by grants from the Netherlands Organization for Scientific Research – NWO ( project number 275-70-021 for the first author, project number 276-70-013 for the second author), which we hereby gratefully acknowledge.
1 For many authors, the notion of a small clause also encompasses verbal predicates, as in I heard him float away (Higginbotham 1983:100). In Slavic linguistics, such is usually not the case, possibly because such constructions are not grammatical in Slavic languages.
2 We use the term propositional somewhat loosely here, since the actual denotation of small clauses is a matter of debate. Thus, Higginbotham (1983) argues that small clauses denote sets of events; Wilder (1992) and Svenonius (1994) distinguish small clauses denoting states of affairs; and Rothstein (2000) proposes that they denote a set of events or a set of states, depending on the category of the predicate. Finally, Maienborn (2003, 2005, 2007) introduces the concept of a Kimian state as the denotatum of copular clauses, which would make small clauses denote sets of K-states. For our purposes, however, it is enough to assume that small clauses have the semantic type ‹s, t›; that is, they are functions from events to truth values.
3 The verb sčitat’ ‘consider’ is the only Russian verb that can appear with both a regular small clause and a za ‘for’-PP. We believe this is because this particular verb also has the nonintensional meaning ‘count’, in which it can in fact combine with a za ‘for’-PP by the mechanism described in section 3. The Serbo-Croatian verb smatrati ‘consider’, which also appears with both a regular small clause and a za ‘for’-PP, can be explained in a similar way. Etymologically speaking, at the core of smatrati is a perception verb motriti ‘watch’, which has undergone a shift in meaning from concrete to abstract—from perceiving a physical object to perceiving a state of affairs. The fact that verbs of visual perception partake in such meaning shifts most readily follows given the noted and uncontroversial fact that ‘‘vision and intellection are viewed in parallel ways . . . most of all vision is connected with intellection because it is our primary source of objective data about the world’’ (Sweetser 1990:38) (see also, e.g., Miller and Johnson-Laird 1976).
4 While Serbo-Croatian AP depictives bear instrumental case, as in Russian, NP depictives must be introduced by kao ‘as’. The ungrammaticality of the za ‘for’-PP depictive in (i) does not therefore affect the argument.
Absolute small clauses and non-PP resultatives are impossible in both Slavic languages under consideration.
5 Example (24b) is due to an anonymous reviewer. To clarify the notation in the examples below, the verb ‘pass’ in English (25c) and Serbo-Croatian (27b–c) can be felicitously interpreted as transitive with a null object (an exam), and (26b) is grammatical if Jesus was getting off the bus or another means of public transportation.
6Kayne (1984) proposes that the double object construction is also built on the basis of a small clause. We will not evaluate this claim beyond noting that this small clause would clearly differ from the small clause complement of attitude verbs, causatives, resultatives, and depictives.
7Beermann (1997:17n5) notes the same fact for German, claiming that the für ‘for’-PP is used identificationally in this case. Beyond observing that ‘as’ has this property as well, she does not discuss the fact that pronouns and proper names do not appear in uncontroversial small clauses. In Marelj and Matushansky 2010, we argue that ‘as’ is not Pred0 either.
8 Unlike Serbo-Croatian and English wh-NPs, Russian wh-NPs are not sensitive to the referential versus predicative use of a noun phrase: the wh-pronoun used for [+human] NPs is always kto ‘who’.
9 The outcome is the same if the complement of for is itself an anaphor, as in (ia), or a picture-NP, as in (ib).
Geronimo mistook Clarissa for himself.
They mistook these posters for very bad pictures of themselves.
We avoided anaphors as the complements of for in order to guarantee that NP2 can at least in principle function as a predicate (for discussion, see section 2.2) and excluded picture-NPs because anaphors in picture-NPs are independently known to be exempt from binding theory (e.g., they are logophors in Reinhart and Reuland 1993).
10 In small clause contexts, the possessive pronoun svoj and the anaphor sebe do not allow the matrix subject as an antecedent to the same degree. While the possessive pronoun svoj favors the matrix subject as an antecedent, the anaphor sebe generally disprefers it (also see (ii), which is a baseline type of example). That binding by the matrix subject is marginally possible is illustrated by (i).
Tristrami smatra Klarisuj dobrom prema sebi%i/j.
Tristram consider.3SG Clarissa.ACC good toward self
‘Tristram considers Clarissa kind/good to himself/herself.’
Tristrami čini Klarisuj dobrom prema sebi*i/j.
Tristram make.3SG Clarissa.ACC good toward self
‘Tristram makes Clarissa kind/good to himself/herself.’
This complication does not cast any doubt on the ability of the small clause subject to bind an anaphor in the predicate, but it might be relevant for other questions, such as how large a small clause really is and whether causative and attitude verbs differ with respect to the size of their complement. Since some researchers (e.g., Kitagawa 1985, Sportiche 1995, Starke 1995) suggest that small clauses are really CPs, such data might be relevant for evaluating their proposals.
11 Two observations are in order here. First, Bailyn (2001) claims that an AP can in fact appear after za ‘for’, and he provides (7b) as an example. However, the choice of the adjective is unfortunate here: in argument positions, svoj can function as an NP meaning ‘a person belonging to the in-group’ (we set aside the question of whether this is a result of NP-ellipsis, which is extremely productive in Russian, or of nominalization). Bailyn’s example is therefore not revealing. The fact that in Dutch and French counterparts of for can appear with APs (though not PPs) with the same set of verbs (Starke 1995) requires further investigation.
Second, as an anonymous reviewer reminds us (see Stowell 1981, Rothstein 1999, 2000), verbs may select for small clauses headed by the predicate of a particular category: for instance, make does not allow PP predicates, while expect and have are only compatible with PPs (and their proform there). This fact does not affect our point, however, since verbs compatible with the ‘‘predicative’’ ‘for’-PP do not differ with respect to the lexical category of the complement of ‘for’. Therefore, an attempt to link the obligatory cooccurrence of the putative ‘for’-Pred0 with NPs to c-selectional properties of the verb selecting the putative ‘for’-PredPs does not render the constraint less arbitrary.
12 The interpretational differences in the directional use of za ‘for’ in the two languages turn out to crucially influence its use with nomination verbs (see section 3).
13 Unlike in Russian, in Serbo-Croatian the default predicative case in primary predication is nominative, as illustrated in (i). While instrumental is perceived as archaic or literary, there is no clear meaning distinction. Smatrati ‘consider’ seems to be among the rare verbs that require their predicative complement to be instrumental.
14 In the Slavic equivalent of the Germanic ‘what for’ (German was fu¨r) construction, za ‘for’ seems to assign no case. (See Danylenko 2001 for arguments that this construction has developed in Slavic and Baltic languages independently.)
what.ACC = NOM for book.ACC you.NOM bought
‘What kind of book did you buy?’
what.NOM = ACC this for book.NOM
‘What kind of book is this?’
Bailyn (1991) suggests that in these cases za ‘for’ is not a preposition but a Pred0, since some instances of Pred0 fail to assign case, leading to case agreement (e.g., in Latin and Icelandic). If this hypothesis is correct, it cannot be extended to the NP1-za ‘for’-NP2 sequences discussed here, since NP2 is marked accusative. If, on the other hand, the čto za ‘what for’ construction is assumed to involve predicate inversion (see Bennis, Corver, and Den Dikken 1998), then the lack of case marking on the NP can be explained by placing it in a specifier position rather than in the complement of za ‘for’ (cf. Leu 2008). As a result, za ‘for’ will have no special status with respect to case assignment in the Russian čto za ‘what for’ construction and can be treated as a preposition there as well. We conclude that the čto za ‘what for’ construction cannot be used as an argument for the existence of a Pred0 lexicalized as za ‘for’.
15 As Sag et al. (1985) note, adverbials of unlike categories can also be coordinated.
AdvP and PP
We walked slowly and with great care.
NP or PP
They wanted to leave tomorrow or on Thursday.
NP and PP
We are open Saturdays, any national holiday, and on alternate Sundays.
Bayer (1996) proposes a semantic alternative to Bowers’s (1993) solution, suggesting that coordination is constrained to combine conjuncts of the same semantic type. Crucially, even if this hypothesis is correct, our argument is not affected, since we expect two small clauses to have the same semantic type.
16 More specifically, Reddy (1979) and Lakoff and Johnson (1980) have convincingly shown the pervasiveness of the conduit metaphor for communication, which crucially relies on the view of communication and perception as motion (see also Talmy 1996).
You can’t get your concept across to the class that way.
His feelings came through to her only vaguely.
17 Examples (71a–b), due to an anonymous reviewer, demonstrate that take can take a propositional complement, and the question arises whether take should be grouped with perception verbs, which are only compatible with stagelevel (VP) small clauses (Higginbotham 1983), or with opinion verbs, requiring individual-level small clauses (Svenonius 1994, Basilico 2003). Despite their clear semantic resemblance to the former class, we treat mental intake verbs as a separate group, because, as we will argue below, ‘for’-PPs are not small clause predicates.
18 The noun president occasionally appears bare with have, get, and want for, as well as with nominate. We have no explanation for this fact, but we note that other nouns are less likely or unlikely to do so.