Chapter One Introduction

1.1 The Anomaly of the Nominal Predicate Licensing in Small Clauses

The main concern of this book is the nominal predicates in the following so-called“small clauses”selected by consider-type verbs, see（1）.

（1）a.I consider John（as）a teacher.

b.The general considered the attack（as）a failure.

c.The mother considered her son（as）a genius.

d.John considered Bill（as）a fool.

The term“Small Clause”（SC, henceforward）“refers to a string of XP YP constituents which enter into a predication relation, but the predicate YP, rather than containing a fully Inflected verb, contains an adjective phrase, noun phrase, prepositional phrase or an uninflected verb phrase”（Basilico, 2003）.Examples in（1）contain a small clause whose subject is predicated by a nominal item.Hence, these kinds of small clause is named nominal small clause（NSC, henceforward）.

Within the standard Case Theory as presented in Chomsky（1981,1986b）, examples like those in（1）pose a problem in terms of the θ-criterion and the Case-filter, which function together as the basic principles determining the distribution of NPs.For the purposes of this research, we shall first assume the following formulation of these principles, see（2）:

（2）a.θ-criterion:Each argument bears one and only oneθ-role, and each θ-role is assigned to one and only one argument.

（Chomsky,1981:36）

b.Case-filter:∗ NP if NP has phonetic content and has no Case.

（Chomsky,1981:49）

Taken together, these two principles impose a requirement that every NP, which is phonetically realized and has semantic content, must be assigned both a θ-role and Case, either directly or via a trace which it binds.

There are three NPs in each sentence in（1）, the main subject, the subject in the embedded clause and the nominative predicate of the SC.Then main subjects are licensed perfectly as the external argument of the main verb and marked nominative case by the Infl（For convenience sake, the Split Infl will not be discussed in detail in this section, and will be considered in detail in later parts）.The embedded subjects are licensed by being marked accusative case as the main verb consider is so-called ECM verb（Exceptional Case Marking verb）, which selects an infinitival IP clause as its complement and case-marks the overt NP at the subject position in the embedded clause with a deficient T head.Semantically, the embedded subjects are assigned a θ-role by the predicates in the small clauses.

The story goes well until now, but when we examine the licensing of the third NP in the construction, the nominal predicate, problems rise. Semantically, there is no constituent that can assign a θ-role to it.It may be argued that as a predicate, this NP does not need a θ-role but rather assigns a θ-role.But there are two reasons to argue against this idea.First, at certain stage of derivation, two NPs, and only two NPs bearing the same categorial properties of [+N, -V], interpretable φ feature set and uninterpretable Case feature, are selected from the Numeration and enter the derivation.Which one should be selected to assign θ-role during the computational process and which should be the one to be in the argument position? It's implausible to assume the computation system select randomly because the sentences in（3）are not acceptable（hence the asterisk）.

（3）a.∗I consider a teacher is John.

b.∗I consider a teacher is him.

It is too hush to say that the ungrammaticality in（3）results from that John or him is pronominal and a teacher can denote one's job and show certain property of a person, and so the pronominals should be selected as subjects, and the rest predicate.The evidence against this argument lies in（1）-b, repeated as（4）here:

（4）a.The general considered the attack a failure.

b.∗The general considered a failure the attack.

c.∗The general considered the failure a attack.

It's obvious that both attack and failure are not pronominals but attack cannot be the predicate no matter it's modified by definite or indefinite determiner.So it's worth probing into the semantic and syntactic rules that license an NP as a predicate.

Second, given that the predicate NP is licensed by the ability to assign θ-role to the subject NP and that the subject NP is licensed by being assigned this θ-role by the predicate, it seems that a licensing circle is formed as in（5）and a problem of circularity emerges.

（5）

With the circularity licensing pattern in（5）, we cannot tell which item licenses the other.If we regard the ability to assign θ-role to the subject NP as the licensing of predicate NP, then the prerequisite is that the subject NP should have been licensed by a licensed predicate.But the subject NP is waiting the would-be-licensed predicate NP to license it.So the ability to assign θ-role should not be regarded as the licensing of the predicate NP but the ability of predicate following the licensing.Therefore it is rational to assume that there is a mechanism in universal grammar to put a teacher at predicate position and John the subject position.We assume the mechanism is closely related to the nature of predication relations and the properties of different kinds of nominals. All the syntactic items like“teacher”, “a teacher”, “the teacher”, “many teachers”, and“all teachers”bear the same category“NP”（or“DP”along with the DP Hypothesis）, but they have various internal structure, semantic references and syntactic behavior as follows.For example（6）:

（6）a.I consider him a teacher.

b.∗I consider him the teacher.

c.I consider them teachers.

d.∗I consider them many teachers.

e.I consider them all teachers.

Sentences in（6）show that the change of modifiers or determiners in NP phrases leads to the difference in grammaticality.We assume here that different determiners change the properties of the nominals and only certain kinds of nominals can be the predicate and trigger the predication relation in SC.This problem will be examined closely in Chapter 4.2.5.

Now, we turn to a more syntactic problem of the nominal predicate, the Case-marking problem.For convenience sake, we repeat（1）as（7）:

（7）a.I consider John as a teacher.

b.I consider John a teacher.

In（7）, the most possible case marker of predicative NP is the particle as preceding the predicate NP.Some linguists like Haegemman（1994）believe that the presence of as is the evidence of the existence of the functional head Agr in small clause, so small clause is a maximal projection of this head.As the head of the project, Agr selects the predicate NP as its complement and casemarks it naturally.But this assumption is not tenable for three reasons.First, the notion that a nominal item enters computation without Case feature and the case is assigned by a case-assigner has been discarded under the framework of Minimalist approach.Case-assignment mechanism within GB framework heavily depends on the notion of government.But the notion has been doubted in MP.Under the framework of MP, case-marking is conducted in a totally different mechanism named feature checking.Second, there are lots of disputes on the internal structure of small clauses.Linguists have yet come to an agreement about whether small clause is a maximal projection or not, which means it's still a question whether there is a head in small clause, let alone the category label of it.So it seems that we should not rush into this case assignment pattern.Third, if we assume that“as”assigns case to the predicate NP, the situation in（7）-b is worse because the only potential case-assigner is missing.But we don't take it as the end of the story.On the contrary, it's the beginning of the story, because it means that the predicate NP is not licensed by“as”in case-assignment patterns but by some other syntactic mechanisms.

But before we probe into the syntactic mechanism licensing predicate NP, we need to tackle with a more fundamentally conceptual question:whether it is necessary for predicate NP being case-licensed since it is not an argument and bears no θ-role.

In the framework of Government and Binding, two views of Case have developed:the original Case Filter proposed by Rouveret and Vergnaud（1980）and Chomsky（1981）as listed in（1）-b rules out caseless noun phrases with phonetic content, regardless of their thematic status.However, this original view was quickly replaced by Chomsky's（1981:336）alternative version which derives the Case Filter from the Theta Criterion and only applies to θ-marked noun phrases.This revision of the Case Filter became known as the Visibility Condition since case marking is treated as a condition making a noun phrase visible for Theta marking.Since most attention has been paid to argument nominals, it is not clear whether the theory extends easily to predicative phrases.However predicative NPs exist substantively not only in SC but also in copular sentences coming after copula“be”.In copular sentence, the overt copula“be”is taken as the accusative case assigner of the predicate NP（Maling and Sprouse,1995）.But it cannot account for the linguistic fact that predicate NP in SC is licensed without any overt case-markers.

Pereltsvaig（2008）pointed out the problems of the revised version of Case theory when studying the intra-structure of Russian copular structures.The first problem with the Visibility Condition is that, being an LF principle, it does not easily relate to the observable morphological case.The second problem is that it is not relevant to case marking on post-copular phrases since presumably post-copular nominals are not marked and therefore need not be marked with Abstract Case either, according to the Visibility Condition.Yet, at least in some languages with morphological case systems - including not only Russian but also Icelandic, Arabic, Korean, Finnish and others - post-copular phrases clearly bear morphological case marking, which a grammar system including the Visibility Condition remains unaccounted for.The solution proposed within the framework of the Visibility Condition is to treat such case marking as“a morphological default in the absence of syntactic case resulting from an independent morphological well-formedness requirement on nouns”（Pereltsvaig,2007:105）.However, this approach fails to account for two uncontroversial facts:first, some case-marking on post-copular phrases is not easily analyzed as morphological default, such as instrumental case in Russian and accusative case in Arabic; second, at least some case-marking on postcopular phrases has been argued to indicate certain syntactic configurations（Lee,1989; Maling and Sprouse,1995; Comrie 1997）.This leads some researchers to challenge Visibility Condition as incorrect or insufficient, proposing to go back to the original Rouveret-Vergnaud formulation that Case serves as licensing mechanism for all and only overt noun phrases（Lee,1989）.

Besides the empirical data of the morphological case found on predicate NPs, we also encounter a theoretical barrier in assuming the caselessness of predicate NPs.Within MP framework, all lexical entities enter computation with features, including categorial features such as [+N, -V], or [+V, -N] and the features determined by categorial features.In this system, nominal items bear φfeatures and Case features along with categorial feature N.And for a nominal, categorial and φ features are [+interpretable], because as a noun it inherently expresses properties like number, gender and person.But Case feature is [-interpretable]because we cannot tell the case of a noun by itself. Case feature must be checked by an interpretable matching Case feature in a functional head for the nominal item to be interpreted at LF and hence the derivation converges.In this system, all nominals need to be case-licensed by checking the [-interpretable]Case feature, including NPs on predicate position.So, theoretically speaking, to exclude predicate NPs from Case Filter which is assumed to explain the distribution of nominals is a great loss in terms of the uniformity and the explanatory adequacy of generative grammar.And the licensing mechanism of the predicate NPs requires a theoretical account.

Therefore, we stick to the previous version of Case Filter and assume that any overt NP, including the non-argument predicate NPs, must be ruled out if it is not properly case licensed.So the gap between linguistic empirical data and linguistic theory we found in NSCs triggered this study and we try to present a solution within the framework of MP.