Zas.gwz-berlin.de

Joan RafelUniversitat de [email protected] The Syntax of Small Clause Predication
Abstract
In this paper I put forward and justify a syntactic configuration that I call Complex SmallClause-structure. I show that this single syntactic structure can explain both the semantic valueand the syntactic behavior of a range of constructions that up to now have been exploredseparately and, hence, proposed divergent analyses among them.
The Complex Small Clause-structure
The syntactic configuration that I want to propose and defend in this article is depicted in (1). Thisis the syntactic configuration of what I call a Complex Small Clause (CSCl).
YPX [CSCl] = EXTERNAL PREDICATION From the bottom up, we can see here that a lexical head X selects an external argument(alternatively, a constituent of lexical material selects an external argument if X appears withcomplements). This external argument is base-generated in the specifier of the projection headed bythis lexical element, that is, in Spec, XP. As usual in a syntactic configuration like this, these twocomponents end up establishing a subject-predicate relationship, which will have to be licensedwithin a functional domain. Typically, the members and the content of each member of thisfunctional domain will be determined by the lexical head X, in the sense that each functionalprojection of this domain will have to be associated with the lexical head of the predicate (X) (seeGrimshaw 1991, Riemsdijk 1998).
In the structure in (1), I only represent the highest extended projection associated with X, which I call YP. The subscript X on the YP-projection indicates the association of this functional category with the lexical head X. The dots between XP and YP, on the other hand, mean that other functionalprojections may also appear between these two projections, but, of course, only if required by thelexical head X.
Now, going back to the external argument of the lexical head X, we can see in (1) that this argument has to be null in a CSCl, that is, it must be either a PRO or a pro. The former will showup if this argument cannot check Case within the functional domain of X, namely somewherebetween the XP-projection and the head Y, or, alternatively, it can only check off null Case.1 Thelatter will appear if it can check off nominative Case within the functional domain of X.
As you may have already noticed, up to this point nothing special has been said in the structure in (1), since the syntactic configuration as described so far actually embodies the syntacticconfiguration of an ordinary predicative domain.
The special thing in the structure in (1), however, arises when we consider the highest extended projection of X, that is, the YP-projection. As it can be observed here, the unusual thing is that aDP-argument appears base-generated in its specifier, i.e., Spec, YP. As indicated by the subscript,we also notice that this DP will have to corefer with the grammatical subject downstairs, which isthe null subject PRO / pro.
Now what this syntactic configuration tells us is that, if this arrangement of lexical and functional categories can be instantiated by some construction in some language, then language ingeneral must permit the possibility for a single extended projection - YP in (1) - to contain twopredicative relationships. In (1), on the one hand, we have the predicative relationship that isestablished by the null subject PRO / pro in Spec, XP and the X´-constituent. I call this subject-predicate relationship the internal predication of the CSCl. As we will see, this internal predicationcan come in two varieties: either as a verbal clause (section 2) or as a Small Clause (section 3).
On the other hand, we also have the predicative relationship that is set up by the DP in Spec, YP and the Y´-constituent. I call this predicative relationship the external predication of the CSCl.
Now the nature of this external predication will determine the status, and hence the behavior, of thewhole construction in (1). At this point, we already know that Y must be the highest extendedprojection of the lexical head X. This means that Y cannot be itself a lexical head, but a functionalelement (or semilexical, grammatical. head (see Corver and Riemsdijk 2001, Rafel 2001). On theother hand, we know that the head of a full clause can only be verbal. Therefore, if the constructionin (1) is headed by Y, which is not a verb, then we can already anticipate that the wholeconstruction in (1) will behave like a Small Clause.2 Hence the name complex Small Clause.
Differently from ordinary SCls, though, here the predicate of this “Small Clause” contains a full-fledged predication. In other words, its predicate is much more complex than that of a regularSCl.3 Hence the term Complex small clause.
What I want to do in the remainder of this article is to discuss the properties of several constructions in order to demonstrate that the syntactic configuration described by the CSCl-structure in (1) really exists. And, importantly, it really exists as a general structure. That is, thisconfiguration is proven to encode the semantic and syntactic properties of various constructions indifferent languages. This means that language must indeed admit the possibility for a singleextended projection to contain two predicative relationships, an idea that, I think, would be worthtaking into account when we intend to explain the semantic or syntactic properties of constructionsthat behave like SCls.
The constructions that I deal with here are divided in two types. Type 1 represents constructions that express an event in progress in Romance and Germanic languages. These 1 This apparent “optional choice” simply responds to the more general controversy surrounding the type of Case thatPRO checks, if any.
2 For the concept of Small Clause, see Stowell 1981, 1983, and for some discussion with regard to this notion, seeCardinaletti and Guasti 1995.
3 Recall that the predicate of a regular SCl is X´, where X is a lexical category (N, A, P). As I point out in section 4.2below, there cannot be “regular” SCls where X is V. I claim that the so-called Verbal Small Clauses are actuallyComplex Small Clauses, where X is V and the (C)SCl-subject is base-generated in the specifier of the highest extendedprojection associated with that V.
constructions are discussed in section 2. Type 2, on the other hand, are complex constructions thepredicate of which is nominal or adjectival. I explore constructions containing the words regard-asand take-for. But within this type I also include resultative constructions in Chinese, although Iclaim that the analysis for the Chinese resultative constructions can also be extended to theresultative constructions in English. This is the topic of section 3. Finally, in section 4, I point outsome general conclusions both for the general Theory of Grammar and for the SCl-Theory that canbe drawn from the CSCl-structure presented in this article.
Complex Small Clauses Type 1: The Progressive
The constructions that are discussed in this section have two properties in common at least. The firstone is that they all respond to the CSCl-structure presented in (1) above. And the second one is thatthey all express an event in progress. In section 2.1, I focus on the so-called Pseudo-Relative inRomance. In section 2.2, I consider the so-called Prepositional Infinitival Construction, which isfound in European Portuguese, in some Italian and English dialects, and in Middle English. But, aswe will see, the nominal version of this construction is also found in languages like German andDutch.
2.1 The Pseudo-Relative
The so-called Pseudo-Relative (PR) is a construction that is used in the majority of the Romance
languages to express an event in progress. An example is provided in (2) for Spanish, (2a), and
French, (2b).4
J´ai vu [PR Jean qui courait.]I have seen to-ACC John that ran.he-IMPERF`I saw John running.´ Before going on, let me just remark that this construction is not a relative clause. There are some arguments that conclusively show that this is so. Here are some of them: (i) In the PR, the that-constituent does not modify the DP, but it rather expresses a situation in which that DP is a participant. This is what allows the whole construction to express an event inprogress.
(ii) The DP can be a proper name in the PR, and, importantly, there is no break in the intonation between the DP and the that-constituent, at least necessarily.
(iii)Differently from a relative clause, the DP can only be interpreted as (or associated with, see shortly below) the subject of the embedded finite verb.
(iv) The tense of the that-constituent must match the tense of the matrix clause only in the PR.
(v) And only in the PR the DP can be extracted leaving the that-constituent behind.
Now, from a semantic point of view, the PR can only express an event in progress. In other words, this construction cannot denote a proposition despite being a CP-constituent. As expected,then, the only type of verbs that will be able to appear in this structure are verbs that are related toevents. If this condition is not satisfied, the sentence becomes ungrammatical. This is what theexample in (3a) shows us.5 4 Constructions like I saw [ John running ] are ungrammatical in some Romance languages like, for instance, Frenchand Italian.
5 In this section I use Spanish data, but crucially the same effects do also hold for the other Romance languages thatpossess this construction.
a. *Vi a [ Juan que sabía francés.] b. Vi [ que Juan sabía francés.] saw.I that Juan knew.he French`I saw that Juan could speak French.´ Notice that here the verb used is saber (`to know´), namely a verb typically linked to propositionalexpressions. The example in (3b), on the other hand, indicates that the verb saber (`to know´) canappear in an ordinary CP-structure, since the inherent semantic properties of this verb are not inconflict with the propositional status of a CP.
There are also some interesting syntactic facts that define the PR. To begin with, it is important to remark that this construction is interpreted as a single constituent, at least in one possible reading.
Therefore, a pronoun like lo (`it´) can resume the whole construction, as illustrated in (4).6 He visto a [ María que corría.] Yo también lo he visto.
have.I seen to-ACC María that ran.she I also it have.I seen`I saw María running. I saw it too.´ Note, incidentally, that this possibility clearly indicates that we are not dealing with a complex DPheaded by the N María in (4), but rather with a “thing.” And this “thing” here is an event.
As far as the assignment / checking of theta-roles and Cases is concerned, we must assume, first, that the constituent headed by the V assigns an external theta-role to an argument base-generated in its Spec, namely, in Spec, VP. This theta-role will be that of AGENT if the V is to run,as in (4). The Case that this argument will check off will be the nominative that is provided by thefinite IP. Now, at this point, we can follow two possible ways: Hyp. 1] The first one is to suppose that the argument that is base-generated in Spec, VP is the lexical DP (Juan). In this hypothesis, then, this is the element that will check off the nominativeCase that is provided by the finite IP.
Hyp. 2] The second approach consists in saying that the argument that is base-generated in Spec, VP is null, and that this is the element that will check off the nominative Case that is providedby the finite IP. Since this null argument checks nominative Case, then it has to be a pro. Noticethat, in this hypothesis, the licensing of pro in the PR does not differ from the licensing of the prothat appears in an ordinary clause, like the one in (5).
Now, if we adopt this second hypothesis, then we must address the question concerning the position in which the lexical DP (Juan) is base-generated in the PR. The claim is that this lexicalargument is base-generated in Spec, CP. This idea is consistent with the fact that this argumentshows up preceding the C that, which is the highest extended projection of the lexical head, namelythe verb, and the fact that the whole construction can be replaced by the pronoun lo `it´ (see (4)).
Notice that this latter fact prevent us from saying that the DP is base-generated in a higher position.
Were this the case, then the whole construction would be expected to behave like a complex DP-structure, contrary to what we have.
The next question that arises from this second hypothesis is how this DP is licensed semantically and structurally. The answer is that it must be semantically licensed by predication. If6 As expected, all the traditional constituency tests can be also successfully applied to this construction. So, forexample, the PR can be clefted, pseudoclefted, the answer to a question, etc.
predication necessarily involves the assignment of a theta-role, then we should assume that this DPgets a theta-role from the C´-constituent, since this C´-constituent is predicated of this DP. On theother hand, the sentence in (6) shows us that this DP is structurally licensed by checking off theaccusative Case that is provided by the matrix verb.
{Lo / La} he visto [ que (pro) corría. ] him / her have.I seen that ran.(s)he-IMPERF`I saw {him / her} running.´ Now this fact is crucial since it immediately allows us to rule out the first hypothesis presentedabove. This is so since, according to that approach, the lexical DP (Juan) would end up checking offtwo structural Cases in the PR. The nominative assigned by the embedded finite IP, and theaccusative assigned by the matrix verb (see (6)). Of course, this goes against the general idea that anargument is frozen in place when it checks structural Case (Chomsky 1995). So, at this point, weare just left with one hypothesis, the second one.
Another interesting fact about the PR is that the lexical subject must necessarily corefer with the null subject pro. In the example in (7), for instance, this condition is not fulfilled. So, asexpected, the sentence is ruled out.
*He visto a [ Maríai que prok corrían. ] have.I seen to-ACC Maria that ran.they The observations provided so far are just part of a battery of arguments that lead us to analyze [CP (CSCl) Juani [C´ que [IP [VP proi [V´ corría ]] ]]] DPi [Yx´ YX . [XP {PROi / proi} [X´ X . ]] ]] Now notice that this structure reproduces the syntactic configuration that is put forward by theCSCl-model presented in section 1 above. The CSCl-model is reproduced here again so we maycompare the general structure, (1), with a specific realization of this model, (8).8 2.2 The Prepositional Infinitival Construction
Interestingly enough, European Portuguese does not accept the PR despite being a Romance
language. Instead, it uses the so-called Prepositional Infinitival Construction (PIC) to express an
event in progress. As we can see in (9), the PIC is formed by a lexical DP, the P a (`at´), and an
infinitive, which can show up inflected, as in (9a), or as a bare infinitive, as in (9b).
Eu vi [PIC os meninos a correrem.] 7 There are two important things that must be pointed out here. The first one is that the PR can be an argument or anadjunct. In the former case, the PR is selected by a lexical head. The analysis, then, would be as shown in (8). In thelatter, it just functions like a depictive SCl. This means that Spec, CP would be occupied by a PRO which would becontrolled by a DP argument. These two versions are also found in the Prepositional Infinitival Construction (section2.2) and in the progressive –ing Construction (section 2.3). The second thing is that the lexical DP in Spec, CP in (8) must corefer with the grammatical subject of the internal predication, independently of the semantic properties of this grammatical subject. That is, the grammatical subject canturn out to be a nonanimate entity or an internal argument. This latter possibility is what we find when the verb in thePR is unaccusative or passivized. Again, these phenomena also apply to the Prepositional Infinitival Construction and tothe progressive –ing Construction. 8 For more details on the analysis in (8) and the analyses that are presented in the remainder of this article, see Rafel2000b.
I saw the children at run-INF-(3P, PL)`I saw the children running.´ The sentences in (10), on the other hand, tell us that the same structure is also productive in some Italian and English dialects, and was productive in Middle English.
him have.I seen at run-INF`I saw him running.´ Not surprisingly, the PIC behaves semantically and syntactically just like the PR. From a semantic viewpoint, then, this construction cannot express a proposition, but only an event. So theargument that was used above to show this very same thing for the PR can be reproduced here againthis time using the PIC. Consider the following contrast: a. *Eu vi [ o João a saber francês.] Eu vi [ que o João sabia francês.] As expected, the verb saber (`to know´) cannot appear in the PIC, (11a), but it can show up in aregular CP-structure, (11b). This indicates that the PIC is a syntactic construction that can onlydenote an event and, because of that, it cannot contain verbs that are not inherently linked to thatontological category.
As far as its syntactic properties are concerned, we must say first that the PIC can also be interpreted as a single constituent in one reading. So it can be resumed by the clitic it or bepseudoclefted. This latter possibility is illustrated in (12).
O que eu vi foi [ os meninos a correr(em). ] what that I saw was.it the children at run-INF-(3P, PL)`What I saw was the children running.´ As usual, we must also suppose here that the specifier of the phrase projected by the infinitive, that is, Spec, VP, hosts the argument that will be assigned the theta role of AGENT by theconstituent headed by the V to run. But, once again, the nature of this argument leads us to considertwo possible ways to proceed.
Hyp. 1] In the first hypothesis, we would say that the argument that is base-generated in Spec, VP is the lexical DP (os meninos). From this viewpoint, this argument would be the one that checksoff the nominative Case that is provided by the IP only when the infinitive shows up inflected. If theinfinitive is bare, then this lexical DP would need to move up into the matrix clause to check offaccusative Case.
Hyp. 2] The second approach consists in saying that the argument that is base-generated in Spec, VP is null. This null argument would be a pro if it can check off nominative Case. This would 9 It is interesting to notice that the DP the deer is not preceded by the P of, which indicates that hunting is a V and, as asuch, it assigns accusative Case.
occur when the infinitive appears inflected. If the infinitive is bare, then the null subject would be aPRO, and presumably would check off a sort of null Case.10 Of course, this second hypothesis needs to tell us where the lexical DP (os meninos) is base- generated in the construction. The answer would be that this DP is base-generated in Spec, PP.
Again, this would be so because of the fact that this argument appears preceding the P a (`at´) andthe whole construction is not interpreted as a complex DP, but as a clause.11 This latter aspectprevent us from assuming that this DP is base-generated in a higher position.
So, according to this second hypothesis, the lexical DP is base-generated in the specifier of the highest extended projection of the lexical head of the construction, namely the verb. Notice that theP a (`at´) is the aspectual element that provides the PIC with its progressive interpretation. Thismeans that this element is a functional head that operates on the infinitive. Now the idea that it isthe highest head of the verbal functional domain is strongly supported by the German and Dutchdata presented shortly below.
The sentence in (13), on the other hand, shows us that the accusative Case that is provided by the matrix verb is checked off by the lexical subject contained within the PIC.
I saw them at run-INF-(3P, PL)`I saw them running.´ Interestingly, it shows us that this occurs independently of the agreement properties of theembedded verb. In other words, the lexical DP checks off accusative Case even when the embeddedIP can provide nominative Case. Again this leads us to adopt the second hypothesis pointed outabove as the right one. Otherwise we would be claiming that an argument can check off twostructural Cases.
Exactly like in the PR, the lexical subject has to corefer necessarily with the null grammatical subject downstairs. Thus, the sentence in (14) is out just because this condition is not satisfied.
*Eu vi [ o Joãoi a {PROk / prok} correr(em). ] Based partly on the analysis that Raposo 1989 proposes for these constructions, partly on the properties that we have seen here, we can say that the syntactic analysis of the PIC is as shown in(15).
[CP [C´ [IP [VP proi [V´ correrem ]]]]]]] [PP (CSCl) os meninosi [P´ a [CP [C´ [IP [VP PROi [V´ correr ]]]]]]] DPi [Yx´ YX . [XP {PROi /proi} [X´ X . ]] ]] As you may have already noticed, the syntactic organization of this construction, (15), alsofaithfully matches the more general syntactic configuration that I have called CSCl-structure, (1).
Before moving on to the English data, let us very briefly consider the German and Dutch [ Jan ] war [ am Schreiben eines Briefes. ] Jan was at.the write a-GEN letter-GEN`Jan was writing a letter.´ 10 See footnote 1.
11 For example, it can be resumed by the clitic it, as pointed out above, and it triggers a third person, singular agreementon the matrix verb when the whole construction occupies the subject position in the sentence.
b. [ Jan ] was [ een brief aan het schrijven. ] Jan was a letter at the write`Jan was writing a letter.´ The elements that make up these constructions are just like the elements that make up the PIC. And,just like the PIC, these structures do also express an event in progress, as we can see through thetranslations into English. Now the only relevant difference between the constructions in (16) andthe PIC is found in the fact that here the aspectual P at is not part of the extended projection of a V,but part of the extended projection of a nominalized V. Since this seems to be the only significantdifference between these constructions and the PIC, it seems plausible, at least in principle, toanalyze the constructions in (16) as indicated in (17). (`Nom.´ means `nominalized version of thePIC´.) Nom. = [PP Jani [P´ an [DP dem [NP PROi Schreiben eines Briefes ]] ]] CSCl = [YPx DPi [Yx´ YX . [XP {PROi / proi} [X´ X . ]] ]] Notice that these constructions clearly show, on the one hand, that the lexical DP is base-generatedin the specifier of the projection headed by the aspectual marker an `at´, namely Spec, PP. And, onthe other hand, that this aspectual element is the highest head associated with the lexical noun, sinceit precedes a DP-projection with an overt D (dem).12 This is strong evidence in favor of the ideathat, in the verbal version of this construction, namely in the PIC, the lexical DP is also base-generated in this position, that is, in Spec, PP, and that the P a `at´ is also the highest headassociated with the lexical head of the construction, namely the verb (see (15)). Now the differencelies in that in the PIC the P a `at´ precedes a CP, the head of which is null.13 2.3 The –ing Construction
An obvious question that arises at this point is whether a similar syntactic configuration like the one
proposed here for the PR and the PIC can also be applied to the progressive construction in Modern
English, in which a suffix –ing appears attached on the verbal head (-ing Construction). An example
is provided in (18) for Spanish, which also admits it, and English.
He visto a [-ing C Juan corriendo. ]have.I seen to-ACC Juan running`I saw Juan running.´ I saw [-ing C John running. ] I can already anticipate that the answer is affirmative, that is, that this construction perfectlyaccommodates to the CSCl-model put forward here. But before presenting the analysis, let me firstremark some properties that show that this construction behaves just like the PR and the PIC.
The first important thing for our purposes here is that this construction does not denote a proposition. So, as we have already seen before for the PR and the PIC, a verb that does not express 12 As David Adger points out to me, the nominal version of the PIC is also used in Irish, as shown in (i).
Chunnaic mi lain na ruith.
saw I John in-AGR (his) running`I saw John running.´ The interesting thing about Irish lies in that the P na `in´ appears inflected. This indicates that there is a pro between thisP and the nominalized verb. That an inflected P is followed by an argumental pro in Irish has been independentlydemonstrated in McCloskey and Hale 1984.
13 The (phonological) null properties of the C must be attributed to the infinitival form of the verb. That is, inPortuguese, as in many other Romance languages, an infinitive is always linked to a null C. So, in this sense, the PICdoes not stand as an exception at all.
an event will not be allowed to appear in this construction. This is the case of the verb to know. Thisfact is illustrated by the already familiar contrast in (19) (cf. (3) and (11)).
a. *I saw [ John knowing the answer. ] b. I saw [ that John knew the answer. ] In this sense, the -ing Construction differs from another construction in English in which the verbalso appears bearing the suffix –ing. This construction, which can be combined with verbs like tohate and to remember, is apparently an ordinary CP and, as expected, denotes a proposition. A pairof examples are provided in (20).
I hate [ {everybody / PRO } telling him what he has to do.] I remember [ PRO having read all these books. ] Some relevant differences between this construction and the progressive -ing Construction are (i) The embedded structures in (20) do not express an event in progress, but a proposition.
(ii) The subject of the embedded constructions in (20) can be a null PRO. This possibility is not available in the progressive -ing Construction. Compare (20) with (21).
I saw [-ing C {John / *PRO} watching the stars.] I saw [-ing C {myself / *PRO} watching the stars. ] (iii)Even though Spanish has the progressive –ing Construction, (22a), it does not possess the propositional construction with –ing, (22b).
He visto a [-ing C Juan corriendo. ] have.I seen to-ACC Juan running`I saw Juan running.´ b. *Odio a [ todo el mundo diciéndole lo que tiene que hacer. ] = proposition hate.I to-ACC all the world telling.him what that has.he that do-INF(intended meaning: `I hate everybody telling him what he has to do.´) Thus, in the Spanish counterparts of the English sentences in (20) we can only find either a that-clause (when the subject of the matrix clause and the subject of the embedded clause do not refer tothe same person), (23a), or an infinitival complement (when the subject of the main clause and thesubject of the embedded construction do refer to the same person), (23b).
Odio [that-clause que todo el mundo le diga lo que tiene que hacer. ]hate.I that all the world him tell what that has.he that do-INF`I hate everybody telling him what he has to do.´ Odio [Inf-clause PRO decirle lo que tiene que hacer. ]hate.I tell-INF-him what that has.he that do-INF`I hate telling him what he has to do.´ This indicates that the progressive –ing Construction and the embedded structures in (20) are indeeddifferent constructions.
Like the PR and the PIC, the progressive -ing Construction can also be interpreted as a single constituent, at least in one possible reading. Thus, the whole (embedded) structure in (24) can beresumed by the clitic it.
I saw [ John running.] I saw it too.
Let us point out now what we know for sure about the assignment of theta-roles and Cases in this construction. First, we know that, as usual, the verb to run in (25) assigns an external theta role(AGENT) to an argument situated in the Spec of its projection, that is, Spec, VP. And, secondly, weknow that a lexical DP contained within the progressive -ing Construction checks off the accusativeCase that is provided by the matrix verb in the example in (25).
On the other hand, we also know for sure that the suffix –ing that appears on the verb is the aspectual marker that provides the construction with its progressive interpretation, and that thisconstruction does not denote a proposition. In other words, it seems fair to think that thisconstruction cannot be an ordinary CP-structure, probably in contrast to the embedded -ingconstructions in (20).
Now, if we put together all the things that we know for sure about the -ing Construction,14 then we are led to analyze this construction the way it is depicted in (26).
[CP (CSCl) Johni [C´ _ing [IP [VP PROi DPi [Yx´ YX . [XP {PROi / proi} [X´ X . ]] ]] Now, as you may have already noticed, the only difference between this construction, on the onehand, and the PR and the PIC, on the other, lies in the morphological nature of the CSCl-head. Thatis, in this construction the CSCl-head is the aspectual suffix –ing. So, as a suffix, it will have toappear at the overt Syntax attached on a lexical element, in this case the verbal head. Differently,the CSCl-head in the PR and in the PIC, namely que and a, respectively, is an unbound element. Soit will be able to show up at Syntax as an independent morphological head. All in all, this meansthat Modern English uses a synthetic version of the progressive construction, whereas thoselanguages that utilize the PR or the PIC make use of the analytic version of exactly the sameconstruction.
2.4 Summary
The specific instantiations of the CSCl-model that have been presented in this section are
reproduced here once again in (27).
→ A single syntactic configuration [CP [C´ [IP PROi correr ]]]]] (analytic) The main properties that characterize these structures are the following: 14 The ones mentioned in the text but also the idea that the verb must be associated with an IP- and a CP-projection; thefact that this construction behaves like a SCl, and just like the PR and the PIC, which do also express an event inprogress; or the fact that the lexical subject can move further up to an A-position: Johni was seen [CSCl ti running. ] Johni is [CSCl ti running. ] (i) The CSCl-predicate (i.e., the internal predication) can be either verbal (Romance, English) (ii) The highest extended projection of X is an aspectual marker: [ Y = Asp ]15(iii) The aspectual marker can be an independent head (analytic version [Romance, German, Dutch, Irish, dialectal and Middle English]) or a bound head (synthetic version [ModernEnglish and some Romance languages]).
Complex Small Clauses Type 2: regard-as / take-for and Resultative
Constructions
The question that arises at this point is whether the CSCl-model presented in section 1 can onlyaccount for the progressive construction in various languages or, differently, this syntacticconfiguration is more productive than that. Well, the answer is that this structure is more productivethan that. I claim that it can also be found in constructions that here I call regard-as and take-forconstructions (section 3.1), and may also be found in resultative constructions (section 3.2).
3.1 regard-as and take-for constructions
The examples of CSCl that here I call the regard-as and take-for constructions are provided in (28).
Once again, let us first start remarking what we surely know about these constructions. To beginwith, we know that the SCl-predicate my best friend and a fool in (28) must assign a theta-role to asubject, just like it does in the sentences in (29).
Johni is [SCl ti my best friend. ] On the other hand, we know that the lexical DP John in (28) checks off the accusative Case that is provided by the matrix verb. This is shown in (30).
I regard him as my best friend.
Now, if we want to make things easy, we can say at this point that the argument that gets the theta-role from the SCl-predicate in (28) and the argument that checks Case within the matrixclause in (30) is exactly the same one, that is, the lexical DP John. Thus, from this perspective, theDP John would be the thematic subject of the SCl-predicate my best friend or a fool, that is, itwould be base-generated in a position following the particle as / for. Later on, it would move up inthe structure in order to check Case within the matrix clause. Hence, this lexical DP ends up in aposition preceding the particle as / for at the overt Syntax. This line of reasoning has actually led tomany linguists to propose what we can call the “traditional” analysis of these constructions. Therepresentation is given in (31).16 15 For arguments in favor of the idea that the C que `that´ behaves like an aspectual marker in the PR, see Rafel 2000a,b.
16 The particles as / for have been taken as `prepositional complementizers´ (see Starke 1995). Even though I also usethis term here, my analysis does not depend at all on the categorial status of these heads.
[CP [C´ as [XP (SCl) John [ my best friend ] ]]] b. [CP [C´ for [XP (SCl) John [ a fool ] ]]] Although the analysis in (31) seems to be the simplest one (even the most logical one), it raises some important problems that should not be ignored for the sake of simplicity. Some of theseproblems are the following:17 (i) Differently from English, in Spanish subjects do not have to appear necessarily in Spec, IP at the overt Syntax (or its equivalent in a SCl). So, for example, in SCls the subject can show upeither following or preceding the SCl-predicate. This is shown in (32).
Tomaron a [ Juan por tonto. ]took.they to-ACC Juan for fool`They took Juan for a fool.´ Despite that, the subject cannot appear in this construction between the particle por `for´ and thepredicate tonto (`fool´) ever. This is illustrated in (33a).
Tomaron a [ Juan por extraordinariamente tonto. ]took.they to-ACC Juan for extraordinarily fool`They took Juan for a big jerk.´ The example in (33b), on the other hand, tells us that the ungrammaticality of (33a) cannot beattributed to some kind of affixation of the particle for onto the predicate fool.
(ii) Following the analysis in (31), we must assume that in the passive sentences in (34) the DP John has moved from the position where it is base-generated (an A-position), to the Spec of theprojection headed by the particle as / for (an A-bar position)18 and, finally, into the subject positionof the matrix clause, where it checks off nominative Case (again an A-position).
Johni is regarded [ ti [ as [ ti my best friend. ]]] Johni is taken [ ti [ for [ ti a fool. ]]] Now the legitimacy of this movement operation is not, by any means, obvious since we obtain amixed [A, A´, A] chain and, according to the generative tradition, this combination should get us anungrammatical output, in contrast to what we have, (34).
(iii)Another question is why the lexical DP in (35a) cannot be assigned structural Case by the P for contrary to what we have in (35b). Furthermore, we may wonder why the D/NP women [3person, plural] in (36a) has to move to a position preceding the particle as if this D/NP does notcheck off the nominative Case that is provided by the finite IP of the matrix clause [3 person,singular].
a. *They took [ for him a fool.]b.
[ Women as engineers ] still surprises some people. (from Emonds 1985: 276) b. *[As women engineers ] still surprises some people.
17 For more arguments and details, see Rafel 2000b, 2001.
18 Since this is not an agreement position nor a theta position in this analysis. Recall, furthermore, that in this analysis asand for are considered `prepositional complementizers´.
(iv) And, finally, in the regard-as case, we happen to have a version in which the particle as introduces a finite clause, as shown in (37b).
I regard John as my best friend.
b. I regard Johni as if hei were my best friend.
In this finite clause we find a subject that, in this example, must corefer with the lexical DP thatappears preceding the particle. Of course, the analysis in (31) does not tell us anything about theobvious relationship between the SCl-version, (37a), and the finite clause-type, (37b).
The position we are at this point is the following. We know that the SCl-predicate my best friend and a fool must assign a theta-role to a subject (see (28)-(29)), and that the lexical DP checksoff the accusative Case that is provided by the matrix verb (see (30)).
But we now know that this lexical DP, first, cannot appear between the particle and the predicate ever, and, secondly, cannot move up into the matrix clause without violating sometheoretical principle.
At this stage, we also know that there are some constructions that contain two subjects, one lexical and one null, within a single extended projection. Now the obvious thing to suppose at thispoint is that maybe a similar analysis can also be applied to these constructions. Were this the case,we would obtain the structures in (38).
DPi [Yx´ YX . [XP {PROi / proi} [X´ X . ]] ]] Now the surprising thing is that by adopting this analysis we can immediately account for theproblems that we encounter by using the traditional analysis. Here are the explanations: (i) The lexical DP (John) cannot appear between the particle and the SCl-predicate because this DP is base-generated in a higher position in the structure, namely Spec, CP. Of course, thesame goes for languages with a relatively free word order like Spanish (see (33a)). But, as we havealready seen, in Spanish, even though we can find the SCl-subject either preceding or following theSCl-predicate (see (32)), and even lexical material between the particle por `for´ and the predicatetonto `fool´ (see (33b)), we can never find the lexical DP between these two elements. This factcould be attributed to a prohibition of inserting an argumental DP-subject into the subject domainalready occupied by another argumental subject, namely PRO, which is, furthermore, coindexedwith, or controlled by, that very same lexical DP.
(ii) In this configuration, the specifier of the projection headed by the particle turns out to be an A-position. This is so because the lexical DP is base-generated there. So this lexical DP will be freeto move further up in the structure to an A-position. Hence the grammaticality of the passivesentences in (34), where we end up with a uniform [A, A] chain.
(iii)The lexical DP cannot get Case from the prepositional C for (see (35a)), as opposed to what we find in structural contexts like that in (35b), because in the derivation this lexical DP neveroccupies a position below that particle.
(iv) The only difference between (37a) and (37b) lies in that the constituent introduced by the particle as is a SCl in (37a), the subject of which is a PRO because it cannot check structural Case,whereas it is a finite clause in (37b), where the subject is a pronoun that can check off nominativeCase.
3.2 Resultative Constructions
Interestingly enough, Huang (1992) proposes an analysis for the resultative constructions in
Chinese that reminds us a great deal of the CSCl-structure that we are testing here.
An example of the resultative construction in Chinese is given in (39).
John [ ba Mary ] ku- [de hen shangxin.] John BA-PREP Mary cry-DE-CLITIC (=OLD V `obtain´) very sad(lit. John to Mary cried till very sad.)`John cried till Mary got very sad.´ In this sentence, we can see that the resultative construction in Chinese is formed by twoconstituents, the ones that appear into brackets. In the first one, we have a DP, Mary, which isintroduced by a particle, presumably a Case marker. In the second one, on the other hand, we have aparticle, which derives from the old verb to obtain, plus an adjectival predicate that indicates thestate in which the DP Mary ends up in.
Now, the analysis that Huang proposes for this construction is the one depicted in (40).19 ku- [Resultative Clause -de Proi hen shangxin.] According to this analysis, the V (cry) selects and theta-marks the resultative clause (“obtain” verysad). After that, the resultative clause (“obtain” very sad) plus the V (cry) select and theta-mark thelexical DP (Mary).
So the question at this point is whether this analysis accommodates or relates, if it does in any way, to the general CSCl-structure that we are using here. Well, the answer is clear cut: It doesrelate to the general CSCl-model since the analysis in (40) is nothing more than a “restructured”version of the CSCl-configuration. The analysis in (40) previous to the restructuring operationwould look like (41).
Resultatives = [CSCl Maryi [ de [ Proi [ very sad ]] ]] [YPx DPi [Yx´ YX . [XP {PROi / proi} [X´ X . Here we only have to say that the CSCl-head, which is de in (41), incorporates at Syntax onto thematrix verb (cry). Nothing else needs to be said.
Just like in other types of CSCl, the lexical subject checks structural Case (accusative) within the matrix clause, and it must corefer with the null subject Pro, which gets the theta-role from theadjectival predicate. As expected, if the subject of the external predication and the subject of theinternal predication do not corefer, then we obtain an ungrammatical sentence. This is shown in(42) (from Huang).
*tak ba fani chi-de [ ti [Prok hen bao] ] he BA-PREP rice eat-DE very full(intended reading: `He ate rice and got very full.´) The natural question to ask at this point is whether the same approach to the resultative constructions in Chinese can also be extended to the resultative clauses in English. I think that, as ahypothesis, this is a plausible idea. We would only need to say that in the resultative constructionsin English the head of the CSCl is null. So the structure of the embedded construction in (43a)would be as depicted in (43b).
19 Huang uses Pro for pro or PRO due to the lack of morphological evidence in Chinese in favor of one or another.
b. [CSCl (PP) the doori [P´ ∅ [= “obtain”] [AP (SCl) PROi [A´ open ]] ]] CSCl = [YPx DPi [Yx´ YX . [XP {PROi / proi} [X´ X .]] ]] As I said, this is a hypothesis that, I think, would be worth looking into.
3.3 Summary
The analyses of the constructions that have been considered in this section are reproduced here once
again in (44).
→ A single syntactic configuration [DP (SCl) PROi my best friend ]]] (analytic) The main properties that characterize these structures are the following: (i) The CSCl-predicate is nominal or adjectival.
(ii) The highest extended projection of X is either a modal marker (as / for) or a relational (iii) This marker can be an independent head (analytic version [regard-as and take-for constructions]) or a bound head (synthetic version [Chinese resultative constructions]).
The Complex Small Clause-structure: Some consequences
In this article, I have put forward the structural model of what I have called a Complex SmallClause, and have applied this model to several constructions in different languages. The (main)constructions that have been examined and the analysis in terms of a CSCl that has been proposedfor each one of these constructions appear in (45).
[DP (SCl) PROi my best friend ]]] (analytic) In this section, I remark some consequences that can be drawn from the discussion presented in thispaper. In section 4.1, some consequences for the general Theory of Grammar are pointed out. Insection 4.2, I outline some consequences for the SCl-Theory. Of course, these general consequencesmust be implemented by the ones drawn by the reader.
4.1 Some consequences for the General Theory of Grammar
We have learned that a structural model, the one put forward by the Complex Small Clause- structure, explains the semantic and syntactic properties of a set of constructions that up to nowhave been analyzed in tremendous different ways.
The constructions examined here have told us once again that every element counts, be it an independent head or a morpheme. So the simplest element can determine the syntax and semanticsof the construction it appears in. For example, we have seen that the particle as that shows up in theregard-as construction is not an “optional” head, that is, the phonological realization of the head ofa PredP-projection, as sustained in Bowers 1993, but the head of a CSCl. So there is an importantsemantic and syntactic difference between the example in (46a), on the one hand, and the ones in(46b, c), on the other.
I consider [CSCl John as my best friend.] I consider [ John as if he were my best friend.] This means that, if there is really a PredP introducing a clause, be it a full clause or a small clause,its head cannot be covert or overt optionally.
The –ing Construction, on the other hand, shows us that the same simple element can also be linked to different semantic and syntactic structures. We have seen that the suffix –ing can beassociated presumably with a plain CP, (47a), or associated with a CSCl-configuration, (47b). In theformer case, the construction has a propositional value, whereas in the latter context it denotes anevent.
I hate [CP people telling him what he has to do all the time.] 4.2 Some consequences for the SCl-Theory
The CSCl-model suggests that predication is the result of a syntactic relationship. We have seen that a full-fledged predicative relationship can be itself predicated of a subject, althoughcertain conditions must be met. The most remarkable ones are (just to recall): (i) The CSCl-subject must be base-generated in the Spec of the highest extended projection associated with the lexical head of the construction (X).
(ii) The CSCl-subject must corefer with the grammatical subject of the internal predication.
This structural configuration is used to express the idea that an entity (DP) is or becomes(progressive and resultatives, respectively) a participant in some sort of event (e) [Type 1] orsituation (s) [Type 2]: A functional (or semi-lexical, grammatical.) element (see Corver and Riemsdijk 2001, Rafel 2001) can be the head of a SCl. This occurs when a subject is base-generated in the specifier of itsprojection. This means that the asymmetry between lexical and functional categories traditionallyassumed by the Small Clause Theory (since Stowell 1981, 1983) does not exist.
The functional properties of the CSCl-head make us expect this head to “look for a lexical host.” It can already do it at the overt Syntax. In this case, we can see that the CSCl-head can lookeither “down,” like in the progressive construction in English [-ing run(n)-], or “up,” like in theresultative constructions in Chinese [eat-de]. But it can also wait and do it after Spell Out. In this case, the CSCl-head shows up at the overt Syntax as an independent head, like in the PR [quecorría], PIC [at work], `regard-as´, and `take-for´ constructions.
The so-called `Verbal Small Clauses´ may just be tokens of the CSCl-model. From this viewpoint, in a full clause the subject would be base-generated in Spec, VP, whereas in a VerbalSmall Clause the subject of the construction would be base-generated in the Spec of the highestextended projection associated with the verb. This is what occurs in the examples of Verbal SmallClause that have been considered in this article, namely the PR, the PIC and the progressive -ingConstruction. But it is also expected to happen in other examples of Verbal Small Clause, typicallyin the so-called Bare Infinitive (BI), (49). ((49a) is in Spanish.) He visto a [BI Juan correr. ]have.I seen to-ACC Juan run`I saw Juan run.´ According to the position adopted here, the BI would be analyzed as shown in (50). Compare (50)with the version of the PIC in which the lexical head is a bare infinitive, (15).
Johni [C´ ∅ [IP [VP PROi [V´ run ]]] ]] CSCl = [YPx DPi [Yx´ YX . [XP {PROi / proi} [X´ X .]] ]] This is a possibility that, I think, deserves to be explored seriously, just like the analysis of theresultative construction in terms of a CSCl addressed in section 3.2 above.
References
Bowers, John. 1993. The syntax of predication. In Linguistic Inquiry 24: 591-656.
Cardinaletti, Anna and Maria Teresa Guasti. 1995. Syntax and Semantics. Small Clauses. London: Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Massachusetts: The MIT Press.
Corver, Norbert and Henk van Riemsdijk. 2001. Semi-lexical Categories. The Function of Content Words and the Content of Function Words. Berlin-New York: Mouton de Gruyter.
Emonds, Joseph. 1985. A Unified Theory of Syntactic Categories. Dordrecht: Foris.
Grimshaw, Jane. 1991. Extended projection. Ms., Brandeis University.
Guasti, Maria Teresa. 1992. Causative and Perception Verbs. Doctoral dissertation, Université de Huang, C.-T. James. 1992. Complex predicates in control. In Control and Grammar, R. K. Larson, S. Iatridou, U. Lahiri, and J. Higginbotham (eds.), pp. 109-147. Dordrecht: Kluwer AcademicPublishers.
McCloskey, James and Kenneth Hale. 1984. On the syntax of person-number inflection in Modern Irish. In Natural Language and Linguistic Theory 6: 143-189.
Rafel, Joan. 2000a. From complementizer to preposition: Evidence from Romance. In Probus 12: Rafel, Joan. 2000b. Complex Small Clauses. Doctoral dissertation, Universitat Autònoma de Rafel, Joan. 2001. As for as / for, they are semi-lexical heads. In Semi-lexical Categories. The Function of Content Words and the Content of Function Words, N. Corver and H. vanRiemsdijk (eds.), pp. 475-504. Berlin-New York: Mouton de Gruyter.
Raposo, Eduardo. 1989. Prepositional infinitival constructions in European Portuguese. In The Null Subject Parameter, O. Jaeggli and K. Safir (eds.), pp. 277-305. Dordrecht: Kluwer.
Riemsdijk, Henk van. 1998. Categorial feature magnetism: The endocentricity and distribution of projections. In Journal of Comparative Germanic Linguistics 2: 1-48.
Starke, Michal. 1995. On the format for small clauses. In Syntax and Semantics: Small clauses, A.
Cardinaletti and M. T. Guasti (eds.), pp. 237-269. London: Academic Press.
Stowell, Tim. 1981. Origins of Phrase Structure. Doctoral dissertation, MIT.
Stowell, Tim. 1983. Subjects across categories. In The Linguistic Review 2: 258-312.

Source: http://www.zas.gwz-berlin.de/fileadmin/material/ZASPiL_Volltexte/zp26/zaspil26-rafel.pdf