Some arguments against some prevalent ideas on specificational sentences

Copular sentences where the copula is flanked by two DPs fall into two types: predicational or ascriptive (John is the physician) and specificational (The physician is John), to use the most widely accepted terms. However, according to the dominant view in the generative transformational framework, this divide, to some extent, is spurious since underlyingly specificational sentences would be predicational. Contrary to this position, I argue that the partition is real, irreducible, and syntactically-based. With this goal in mind, I discuss and reinterpret some well known data and present some new ones.


Introduction
Copular sentences of the form DP copula DP are of a great interest for the theory of syntax.This is due to the fact that they do not present an unsaturated predicate head to drive the derivation.This circumstance seems to suggest that they, in a sense, must respond to the very workings of syntax to a much greater extent than the sentences built around an unsaturated lexical head.
Copular sentences where the copula is flanked by two DPs fall into two types: predicational or ascriptive (John is the physician) and specificational (The physician is John), to use the most widely accepted terms.Other additional types have been proposed (Higgins 1979, Declerck 1988, and Den Dikken 2005 for a review), but in general this major basic partition is assumed.This is true even for those considering that the divide is between predicational and equative types (Dr.Jekyll is Mr. Hyde), specificational sentences being equative (Heycock & Kroch 1999, 2002).And this is true too for copular sentences presenting other expressions than two DPs around the copula: to focus on sentences with two DPs is only a limitation of this paper.
A canonical specificational sentence (The best jogger in the group is Peter) can be intuitively described as contributing a value for a variable where the value corresponds to the postcopular DP and the variable to the precopular one.In a specificational sentence, the precopular DP is not referential: it is the postcopular DP that says who or what the referent is.In a canonical predicational sentence, by contrast, the precopular DP picks out a referent, and the postcopular one assigns a property to it (Peter is the best jogger in the group).In this paper, I will argue that the divide between specificational and predicational sentences is real and irreducible, arising from sheer structural possibilities that yield different convergent outputs at the Conceptual-Intentional interface.Specifically, in my proposal both types of sentences differ from the start of the derivation: predicational sentences have a Predicate layer with an external argument whereas specificational ones do not, their precopular DP being directly merged into the Spec of TP.This view stands against the dominant one in the transformational approach where it is sustained that a specificational sentence (SS) is the result of the raising of the predicate and, therefore, underlyingly predicational.See for instance Williams 1983;Partee 1986;Heggie 1988aHeggie , 1988b;;Fernández Leborans 1991-92, 1999;Heycock 1994;Moro 1997Moro , 2000Moro , 2007Moro , 2008;;Mikkelsen 2002Mikkelsen , 2005;;Den Dikken 2006, etc.Also in contrast with a prevalent idea appearing either independently of the previous one (Heycock & Kroch 1999, 2002;Alsina 2004) or in combination with it (Mikkelsen 2005), I will briefly suggest that informationstructure properties are not crucial but orthogonal to these sentences (see 2.2.9).Informational properties in these accounts play their role either at a post-syntactic level (Heycock & Kroch proposals) or by means of topic/focus features integrated in syntax (Mikkelsen 2002, 2005, Alsina 2004).
In section 2, the main shortcomings of the dominant view are presented.Section 3 contains the main ingredient of my proposal, namely the structural distinction between a predicational and a specificational sentence together with a preliminary formulation of a sort of algorithm able to predict when, given a copula flanked by two DPs, a specificational reading will arise.

Some limits and a caveat
In this paper, I will only deal with copular sentences where the two DP are definite and either D has a NP complement or the whole DP consists of a proper noun.This means that I will not deal with pseudoclefts or with identificational sentences as That is Susan (truncated clefts according to Mikkelsen 2005) although I consider them specificational, too.In the same vein, I will exclude from consideration paracopular verbs even if they might allow a specificational reading in some languages.
I will always use the numeral subscripts in DP 1 and DP 2 , to refer to the DP appearing in the first or second superficial position, respectively.

Shortcomings of the predicate raising view
Although the predicate raising view on copular SSs can be considered the received view, at least in the transformational approach, the difficulties this proposal presents are numerous.I list the majority of those I have been able to detect below and that in one way or another will reappear throughout this paper.I advance, however, that for the sake both of space and relevance I will concentrate only on the first eight points.
The predicate raising view fails when trying to (i) Identify the syntactic source of the distinction between ascription and specification (see 2.2.1, 2.2.2). (ii) Raise the DP predicate without costly stipulations in order to avoid a violation of the locality conditions for movementin the nonsymmetrical views of the predication (see 2.1) (iii) Take into consideration that in SSs across languages one of the two DPs featuring in the construction seems to fail to be formally licensed (see 2.2.4).(iv) Offer a sound account of the agreement patterns found in these sentences across languages (see 2.2.5).(v) Account for the impossibility of the specificational reading for [DP 1 ho 1 copula DP 2 ] in Catalan and other Null Subject Languages (NSL) where ho, of course, is replaced with the corresponding form in the language at stake (see 2.2.6).(vi) Account for the fact that inversion of DP 1 and DP 2 around the copula in a specificational sentence does not necessarily yield a truthconditionally equivalent predicational sentence (see 2.2.7).(vii) Explain the markedness of SSs vis-à-vis copular predicational sentences (see 2.2.8).(viii) Give some hints about the conditions for a sequence DP 1 copula DP 2 to acquire a specificational reading (see 3).All the previous difficulties except (ii) are inherited by Moro's version of the predicate raising view in which the derivation starts with a symmetric small clause, the so called "bare small clause".This symmetrical view of the copular predication relationship presents even further difficulties that I will not tackle given the limits of this paper.I want to point out, however, which ones are the main ones, namely (ix) To explain how the assignment of (logical) subject and predicate roles proceeds.(x) To enlighten how it is that symmetry, a condition apparently incompatible with the workings of syntax, has to be exceptionally admitted as a first valid step in a syntactic derivation.Before proceeding with the shortcomings (i)-(viii), it is convenient to bear in mind some different variants of the predicate raising view.

Specification as predicate raising. Proposals
Long before the influential book by Moro (1997), The raising of predicates, the idea that SSs are inverted predicational structures had been proposed by Heggie (1988) and even before by Williams (1983) andPartie (1986).Having said this, I have to note that, to my knowledge, Moro has never appealed to the very distinction between specificational and predicational sentences in his theory.However, given that SSs will be analyzed as sentences with predicate raising in his approach, it seems perfectly legitimate to present his analysis as a proposal to account for the divide between specificational and predicational sentences.This is normal practice, on the other hand, as the contributions by Mikkelsen, Heycock and Kroch, among many others, testify.Let us then see Moro's proposal together with that of Heggie.Comparing both proposals, a symmetrical predication can be observed in both of them: in Heggie's analysis it is contained in the DPpred and in Moro's view it is within the so called "bare" Small Clause (SC).The proposals in (3) and (4) differ, however, in the landing site of the raised predicate: according to Heggie, it is the Spec of CP whereas according to Moro, it is the Spec TP, which amounts to saying that in the first proposal the landing site of the raised subject in a predicational sentence is not the same as the final position reached by the predicate in a specificational sentence.This difference is worth mentioning since, among the predicate raising proposals, that of Heggie seems to be the only one in the literature attempting to provide a syntactic foundation for the divide between specification and ascription (or predication).Shortly, after presenting the analyses proposed by Mikkelsen and Den Dikken, I will take up this issue again.Mikkelsen (2005) (5) a. Specificational Brian is the best candidate.
[where R means relator ] The common trait here that opposes both analyses to the pair of the two previous ones is that the predication is not conceived as a symmetrical relation in a bare small clause.In (5) as much as in (6), within the PredP and the RP, respectively, the predication relationship is asymmetrical since it is mediated by a functional element (Pred and Relator, respectively).A more salient difference between (5) and ( 6) concerns the predicate that raises to Spec TP in specificational sentences: for Mikkelsen, in a sentence as The winner is John, the DP 1 the winner, on its own, constitutes the predicate merged in the first step of the derivation as complement of Pred 0 whereas for Den Dikken, this DP is not the complement of the Relator driving the sentence but the complement of a Relator embedded in a much more complex Pred, which amounts to treating simple SSs of the sort I am interested here as concealed pseudoclefts.Note that the fact that this complex Pred raises to the Spec TP is what justifies the consideration of Den Dikken's proposal as a variant of the predicate raising view.
Setting aside the issue of unifying the treatment of pseudoclefts and simple copular SS, which exceeds the goals of this paper, the proposals considered in the preceding paragraphs are really very close, for they uniquely differ in the (a)symmetry of the predicate relationship and the landing site of the predicate that raises.The first two, (3) and (4), embrace the symmetrical view and the last two, ( 5) and ( 6), the asymmetrical one.If I am right in my judgment that the symmetrical view has the extra problems mentioned at the beginning of the second section in (ix)-(xi), one will have to incline to the latter two.However, even the proposal by Den Dikken would have, in a sense, to be moved towards the symmetrical side of the debate for it presents a nondirectional view of the predication, where non directionality means that the predicate can be the Specifier of Pred 0 (the Relator head in Den Dikken approach) and, accordingly, the logical subject can be the complement of this head.Given the syntactic indeterminacy this introduces when dealing with the definition of predicate-argument relations or, in other words, once we realize that Den Dikken's proposal would be affected by the deficiency listed below (ix) at the beginning of the current section, which is an important shortcoming to my view, we are left with Mikkelsen's proposal alone.Although Mikkelsen's proposal in ( 5) is the only one escaping from the problems symmetry introduces in syntax, it is, however, unable to explain how the raising of the predicate respects locality (see point (ii) in the list opening section 2).Certainly, the stipulation Mikkelsen (2005: 179-184) introduces, what she calls clumping, would be unable to deal with the variation found across languages for it requires that unom, EPP and utopwhere 'u' means uninterpretable features; 'nom', nominative; and 'top', topic clump together on T and, accordingly, are checked at the same time by a DP that possesses all these same features.The requirement thus precludes the alleged DP subject , Ingrid Bergman, unmoved from its original position in (5), from intervening the relation between T and the the alleged DP pred , the actress, in its original position in (5).This idea, however the theoretical details of implementation might be formulated, is inherently unable to deal with the pattern of agreement in SSs in Romance NSL and Dutch, where the nominative (and agreeing) DP is the DP subject in postcopular position.In these languages, therefore, SSs do not clump these features together in the alleged DP pred so that the DP subject would indeed act as an intervener blocking the raising of the predicate: the prediction of the analysis being, contrary to fact, that SSs in these languages could not exist.In conclusion, if a symmetrical view of predication (or non directional à la Den Dikken) is assumed, one has to cope with problems (ix)-(xi) in the opening list of the current section.By contrast, if an asymmetrical view is propounded, one must resort to really ad hoc and language specific stipulations.
Once the (a)symmetry issue of the predicate raising analysis has shown important weaknesses at either stance, we can concentrate on even more poignant aspects of the predicate raising theory.
Assume for the sake of the argument that an information structure explanation for SSs, in isolation or in combination with a syntactic mechanism, is devoid of explanatory capacity, something that is not difficult to show as we will suggest below (see 2.2.9).Would it then be possible to build on syntax to derive the difference between a specificational and a predicational reading?And, otherwise, where does an specificational reading emerge from?2.2.Is there a syntactic foundation for the divide between specification and ascription?
As advanced supra, the answer to the question heading this section would be negative for all the proposals in the same vein as those examined up to here except the first one, that of Heggie -I will not consider that of Den Dikken given the limits of this contribution.There is a problem, however, with Heggie's proposal, as has nicely been shown by Mikkelsen (2005): the different syntactic structure for the specificational (vs. the predicational) reading in (3a) is only suitable for topicalized predicates that are predicational in interpretation.This being so, the only attempt to provide a syntactic foundation for the divide between specification and ascription must be deemed as failed.Danish data adduced by Mikkelsen are unequivocal in this respect.Let us consider them.

Non-specificational predicate topicalization. Evidence from Danish
Consider the following Danish utterance: (7) Den hjeste spiller pa holdet er Minna the tallest player on team.DEF is Minna 'The tallest player of the team is Minna / Minna is the tallest player of the team.'This string of words is ambiguous: it can receive a specificational or a predicational reading for either of which there is plenty of evidence that a different structure has to be assigned: (8) a. Specificational structure At first glance, one could believe that the key of the difference lies in the different landing site of the raised predicate.That would be wrong.This difference is precisely the only one that can be considered irrelevant for the issue at stake here.As Mikkelsen herself has noticed what is relevant here is that the position of the DP 1 in a specificational sentence has to be the same as the position of the subject in a canonical sentence.As Danish is a V2 language and it remains an open question whether in a subject-initial V2-clause, the verb is in C or T and, accordingly, whether all matrix clauses are, respectively, CP or subject-initial V2-clauses are smaller than that and are TP, it could be that in (8a) the landing site for the raised predicate was the Spec of CP and not the Spec of TP, as arbitrarily assumed in (8a).This would occur, of course, if the uniform CP analysis was finally the correct one.
Once this point has been cleared, the question remains as to what it is that structurally makes the difference between a specificational and a predicational reading in the predicate raising framework.And, of course, the difference has to be looked for below T. There, one can observe that the allegedly logical subject is lower in a specificational sentence than in a predicational sentence: in the first case it has not been moved at all whereas in the second case it has been raised to Spec TP.That the evidence for this distinct positioning of the DP 2 appearing in a specificational sentence and in a sentence with a topicalized predicate is compelling can hardly be seen from an example as (7).It suffices, however, to take into account in either case the behavior of a pronominal DP 2 and its position when the sentence is negative in form to conclude that, in a predicate raising approach to SSs, something like (8a) and ( 8b) is needed for SSs and predicate topicalization sentences, respectively.Certainly, in an SS, a DP 2 would bear accusative case, the default case in Danish, and would appear after negation, a functional projection assumed to be between vP/PredP and TP: two sound reasons to justify the low (and unmoved) position of DP 2 .In a predicate topicalization sentence, by contrast, the DP 2 would be nominative and preceding negation: two reasons for locating it in Spec TP.
The contrasts involving the new data in (11), namely that of (11aiv) vs. (11biv) and that of (11av) vs. (11bv) could be given an account in the framework of predicate raising in terms of reconstruction, an operation that in Danish would be obligatory even for expressions containing a NPIwhich is not the case in English, for instance.Thus, topicalization, being an instance of A'-movement, carries reconstruction whereas the raising operated in an SS, namely that of the DP pred to the Spec of TP, the canonical position of subject, does not.In this way, ikke, the negative expression located between VP and T and responsible for the licensing of the NPI nogen via c-command, would legitimate a NPI in the DP 1 of a sentence with predicate topicalization because, once this DP 1 is reconstructed as the complement of Pred 0 , it is in the right c-command configuration.A NPI inside the DP 1 of an SS, by contrast, could not be licensed because of the absence of reconstruction for expressions raised through A-movement and, accordingly, of the appropriate c-command configuration.
The resort to reconstruction only in the case of the A'-movement of predicate topicalization could be extended in a natural way to explain the contrast between (11av), where reconstruction does not apply, and (11bv), where it does.Surprisingly, Mikkelsen (2005) did not explain this contrast exactly that way: " […] which means that at no point in the derivation is the intended binder in the requisite position.Under the present proposal, we can thus understand the contrast between (2.57) [(=11bv)] and (2.61) [=(11av)] in terms of the different position of the intended binder.In the former it is in subject position, and hence able to bind the reflexive.In the latter it is not, and the reflexive goes unbound."(Mikkelsen 2005: 26) Notice, however, that contrary to what is stated in the quotation, the reflexive inside the precopular DP in an SS has the intended binder in the right c-commanding configuration before this DP raises: recall that the precopular DP is, in this analysis, the complement of Pred 0 and, therefore, the DP subject in the Spec of PredP binds it.This is, however, a minor problematic detail in Mikkelsen's proposal.Perhaps not so minor would be the effect of adopting the view that subjectinitial V2-clauses are CP on the alleged A status of the movement that raises the predicate in an SS.
Be that as it may, the analysis by Mikkelsen (2005) correctly predicts the differences between predicate topicalization and SSs listed in (11), which Heggie's analysis would be unable to deal with.

Why do not we find predicational specificational readings?
The nice result of the proposal by Mikkelsen (2005) differentiating predicate topicalization and SSs does not constitute by itself an answer to the question which entitles the section 2.2: Is there a syntactic foundation for the divide between specification and ascription?Paradoxically, Mikkelsen 2005, correcting Heggie 1988, spoils the prospect of presenting a syntactic explanation for the divide between specificational and predicative readings.In connection with this, we ought to note that the structures in (11a) and (11b) do not present an SS and a canonical predicational sentence, respectively, but an SS and a predicational sentence with a topicalized predicate.One must return to (5), repeated below in (12), to see how Mikkelsen (2005) analyzes the minimal pair consisting of an SS and the corresponding canonical predicational sentence, so to speak.
(12) a. Specificational The actress is Ingrid Bergman.b.Predicational Ingrid Bergman is the actress.Some arguments against some prevalent ideas specificational sentences 119 Observe that the only difference between the members of the pair lies in which of the two DPs raises to Spec TP.Assuming that this position is topical in nature, an SS reading will arise when the DP predicate reaches it and the postcopular DP, unmoved from its initial position, receives a focus interpretation.In contrast, it is the predicational reading that emerges if the two DP move around the copula the other way round and, consequently, the logical subject ends as the topic of the sentence and the predicate as the focus.As we will see in more detail below (see 2.2.8 and 2.2.9), this difference, however, can hardly account for SSs given that the same informational structure, if obtained by means of other structures, does not yield the specificational reading but the predicational one.We are left, therefore, without a sufficient understanding of the source of the difference.
The problem is even more grave when one tackles a question surprisingly unnoticed: there do not exist predicational specificational readings, contrarily to what should be expected if we assume that predicate-argument structure is defined in the first of the three main layers of a sentence (vP/PredP, TP and CP), that is in the PredP in a copular sentence.In effect, that SSs can not be at the same time predicational is something in need of an explanation since it contravenes an implicit core law of the computational system of human language, namely the preservation of the theta structure (or predicateargument relationships) through all the derivation of a sentence.
Taking the previous point seriously, one can not admit that a specificational sentence has the same theta layer as the "corresponding" predicational one.If this is correct, the basic proposal in this paper deserves special attention for it separates both cases from the start of the derivation, as we will see in detail in section 3.

Non specificational left-dislocated DP predicates in Catalan
To some extent paralleling the predicate topicalization in Danish, Catalan and other Romance NSL present clitic left-dislocated DP predicates that, although they are in general relatively marked in comparison with left-dislocated NP predicates (( ?El) president ja no ho és(,) en Joan: lit. the president ADV not CLITIC is the Joan), are always predicational.Interestingly, the idea I propound here can also shed light on their differential behavior with respect to the precopular DP of an SS.Let us take a quick look at them.
Left-dislocated DP predicates can host a reflexive.[specificational] the M.SG his own i cook is the M.SG Aleix Danish and Catalan differ regarding NPI.In Catalan, a NPI can appear not only in object position but also in subject position and, accordingly, it can appear in the precopular DP of an SS.Nonetheless, in a left-dislocated predicative expression they result marginal at best ( ?Cap nin no ho és, es nostro fill (Majorcan Catalan): lit.no child NEG CLITIC is the M.SG POSS.SG 1PL son).As will be shown below (2.2.7), the wellformedness of NPI in subject position allows a kind of SS that is special because its "inverse" predicative version is not truth-conditionally equivalent to the specificational one.
Case and agreement work also differently in Danish and Catalan SSs and this, interestingly, allows us to complement the list of special characteristics of this kind of sentences.Let us see how.

The lack of formal checking for either DP 1 or DP 2 in SS
As mentioned before, in Danish the postcopular DP in an SS bears accusative, the default case in this language.English and French, for instance, are aligned with Danish in this regard.Considering this to be a sign that this DP is not formally checked, an interesting contrast arises since in Catalan, as in Romance NSL in general, and in Dutch, it is the non agreeing precopular DP that has to be deemed unchecked.Thus, SSs seem to be characterized by having either DP 1 or DP 2 unchecked, depending on the language: DP 1 in Romance NSL and Dutch, and DP 2 in English, Danish, etc.Although expressed either by absence of agreement or case default, it seems natural to interpret the exceptional behavior of this construction in this regard as showing that the game is not between predicate and argument(s), which is the domain where the formal licensing by agreement and/or case occurs.Furthermore, the coincidental behavior of Romance NSL and Dutch points to a certain degree of arbitrariness in the selection of the DP that will remain unchecked since there is no known parameter unifying these languages.A quantificational structure like that put forward below (see 3.1) seems again to be able to cope with this exceptional characteristic in a natural way since the relation between the DP 1 and the DP 2 is not a predicateargument relationship, but a quantificational one mediated perhaps by something akin to a feature specification algorithm along the line of that presented in section 3.2 rather than by case/agreement checking.

Agreement patterns
The agreement patterns found in SSs across languages are at first glance complementary with the facts reviewed immediately supra.In this connection, the received view regarding agreement in SSs states that in Romance NSL, copular agreement is with the postcopular DP (the logical subject, according to the predicate raising approach) and with the precopular one elsewhere.The sentences in ( 16) and ( 17) above illustrate this.
But Dutch undermines this, so to speak, parameter-based approach to agreement in copular SS since, as previously mentioned, Dutch presents the general pattern found in Romance NSL, namely agreeement with the DP 2 .
'I believe that the biggest problem the children are/is.' [Den Dikken 2006: 96] And the following unnoticed Catalan data, which, to my knowledge, can be replicated in all Romance NSL, are further counterexamples to the approach mentioned.
(19) Els meus amics són aquesta gent d'aquí.the m.pl poss.pl1p.sg friends be pres.3pldem fem.sg people of here As the gloss indicates, when the DP 1 is plural and the DP 2 is formally singular, but collective, the agreement is with the precopular DPin informant subjects using the singular number to agree with the collective, of course.It seems to me that this kind of example corroborates the idea defended here that in SSs the interplay among the two DPs and the copula is not the same formal interplay that legitimates predicate-argument relationships.This is even emphasized when one takes into account that all the agreement facts in SSs in Romance NSL, including the one illustrated in ( 19), fall into place resorting to the feature specification algorithm to be put forward in the next section.

There is no clitic resuming DP 1 in an SS: DP 1 is not a topic
Consider (20a) and (20b): they are (slightly) marginal as predicational sentencesimproving a lot when the postcopular DP is right-dislocated -, but absolutely impossible as specificational sentences: (20) a. *El guanyador ho serà ell/en Joan. the M.SG winner CLITIC be.FUT.3SG he 'He will be the winner.' b. *El problema ho ets tu. the problem clitic be.pres.2sgyou nom Recall that we have seen through the two previous sections that in SSs in Romance NSL, the general case for the DP 1 is to remain unchecked for agreement purposes: DP 1 does not agree with de copula.How is it, however, that ho, a perfectly suitable clitic for the DP 1 , can not resume it?The fact that it does not allow agreement nor the clitic ho, makes the DP 1 in an SS really exceptional: to my knowledge, except for topicalization in Portuguese, this is the only known case in Romance NSLs of a "core" non-focalized DP in the front of a sentence that occurs with no kind of resumption (inflection or clitic).
This very fact itself suggests at least two important points that should be stressed since they clearly show the weakness of the received view on SS.The first one is that this DP can not have originated lower in the structure, in the way it is in the predicate raising approach where it arises right at the bottom of the tree, as the complement of Pred 0 .And the reason is that only the phrases with a low origin in the structure can be resumed (Boeckx 2003).On the other hand, the absence of a clitic points out that the DP 1 is not a real topic: it is a well known fact that topics in Catalan always carry inflection or clitic as internal resumptive mechanisms.Thus, left-dislocated DP predicates acting as topics obligatorily carry clitic resumption, although naturally the sentences including them are predicational (Es teu professor no ho sere més: lit. the MSG your MSG teacher NEG CLITIC be FUT3SG more; I'll no longer be your teacher).Summarizing, the absence of clitic for the precopular DP of SSs in Romance NSL suggests both that DP 1 originates high in the structure and that it is not a genuine topic.Again both aspects can nicely be accounted for under the analysis propounded here (see 3.1) and, accordingly, constitute recalcitrant problems for the received view in the transformational framework.

Irreversible SSs
One of the arguments often adduced in favor of the predicate raising view is that the result of merely inverting the DPs around the copula in an SS is always a (predicational) sentence with the same truth value than the original one.This fact is argued to point at a common derivational origin for both sentences.Nevertheless, this generalization is not free of serious problems.For space reasons I can not deal with the question thoroughly and I will limit my explanation to some data coming from Catalan and Catalan Sign Language.
As mentioned before, in Catalan a NPI can appear in subject position.This possibility is at the basis of SSs like the one in (21).Interestingly, this sentence is not truth-conditionally equivalent to ( 22 This difference, on the other hand, is exceptional in the sense that nowhere in Catalan, where postverbal subjects are possible, a similar behavior can be observed: a sentence with a negative subject has the same truth value independently of whether it occurs preverbally or postverballyin this last case the clausal negation no is necessary to legitimate the postverbal NPI: Ningú ha arribat is equivalent to No ha arribat ningú: Nobody has arrived; Ningú ha dit la veritat is the same as No ha dit la veritat ningú: Nobody has told the truth).Again, I take this exceptional behavior as evidence that the derivation of an SS does not share a common departure point with the derivation of the apparently equivalent predicational sentence.In other words, the facts under ( 21) and ( 22) suggest the correctness of an approach like that presented in detail in section 3 where specificational and predicational sentences are built differently from the start.
Let us now consider how the way SSs are built in Catalan Sign Language corroborates the previous suggestion.In this language, it is impossible to invert the form of a specificational sentence which, interestingly, consists of a precopular nominal, instantiated as a rethorical question with the corresponding non manual marker for it, followed by the nominal expressing the value accompanied at the same time by a head nod as non manual marker.Notice that their form, and specially the interrogative form on the first nominal, is pointing much more to a view like the defended here than to the received view.Moreover, the fact that, among the optional particles preceding or following the second nominal, the one that more unequivocally marks the focus is excluded suggests that thinking of SSs as topic-focus articulations is a wrong move.

The markedness of SSs
As has been noticed in the literature, SSs in comparison with copular predicational sentences present a much more restricted rank of possible forms across languages.Setting aside cases with omission of either DP 1 or DP 2 , which will enlarge the result enormously, consider the following examples: (23) En Joan és el president (=En Joan és el PRESIDENT) [predicational] 'Joan is the president.'Although these examples in no way exhaust the possible forms even respecting the limitation of no DP omission, they offer a clear indication of what is at stake here, namely the markedness of SS.In Catalan,only (27) and possibly ( 28), whose analysis I can not undertake here, are SS.How is this possible?2.2.9.Some very short remarks on the informational view on SS Some authors (among them Heycock & Kroch 2002, Mikkelsen 2005: 9.2.5) have appealed to a fixed topic-focus articulation to explain the markedness of specification vis-à-vis predication.Thus, it has been observed that predicational sentences can carry focus on either DP while SSs admit it only on the postcopular DP, hence the contrast supra between ( 23)-( 24), both predicational, and ( 25)-( 27), where only ( 27) is an SS.But I am highly skeptical about the validity of this approach for it appears unable to explain neither the reason for this information structure rigidity nor the reason why sentences with the same topic-focus articulation as a canonical SS do not yield specificational readings.For the sake of concreteness, we can ask first why a focus on the precopular DP of an otherwise SS can only be predicational, as (24) shows, and, second, why are not (25) above or (29) below specificational as ( 27) is.
Contrarily to the approach being criticized here, which lacks an answer, my proposal, admitting it is in need of a formal elaboration, seems to be well equipped to offer an explanation for these facts.Thus, intuitively, I would say that (25) can not be an SS because an advanced focus contains its own whole quantificational structure, which precludes the postcopular DP to enter it; and, similarly in (29) where, additionally, the intended value, necessarily in an A position, would be dependent on two operators so to speak, the advanced focus itself and el president, which, of course, is impossible.
Finally, recall that there is an even more important difficulty for the informational view when it is associated, as in general in the theoretical approach (but see Alsina 2004), with the predicate raising approach.As mentioned before (see 2.2.6), given this combination of theoretical points, one would expect that examples such as those in (20) were wellformed SSs, contrary to fact.

For an alternative proposal in the realm of quantification
In my view, what distinguishes specificational from predicational sentences is the syntactic structure they have.The approach presented here complements the structure in question with a sort of specificational algorithm aiming at predicting both when a sequence of the form DP copula DP will yield a specificational reading, and which DP will agree with the copula in Romance NSL.

The structure
My proposal completely differentiates between a specificational and a predicational sentence: The winner 1 is John 2 .
The structure in ( 30) is non-predicational or non-ascriptive because of PredP appearing without an external argument.Complementarily, the predicational character of ( 31) is understood as the outcome of a PredP with an external argument.Moreover, the structure in (30) allows us to directly express what an SS is, namely a quantificational structure where DP 1 contributes an operator (D) and its restrictor (NP) to create a special kind of variable, the DP 1 itself, for which DP 2 is the value.In this connection, notice that DP 1 being directly merged into Spec TP, is neither an argument, nor a predicate, both characteristics being syntactically expressed in the theta layer, but an A' element.
As can be seen in ( 30) and ( 31), I will assume, without further discussion, that the copula is not a verbal element but the exponent of T 0 .If the copula is not a verb, there is no reason for adopting the view that in a copular sentence the complement of T is verbal (Svenonius 1994).PredP is a functional projection whose head can be phonetically empty (Bowers 1993(Bowers , 2001)))).
By assigning the structure in (30) to specificational sentences, an insightful understanding of their markedness obtains (see 2.2.8).Setting aside copular SSs in which the expressions flanking the copula are not DPs, we could say that a specificational reading is the result of directly merging a convenient DP to the T projection of a PredP that is lacking an external argument.The structure underlying copular SSs is, therefore, akin to unaccusative non copular sentences in that both have a predicate layer without an external argument.Unaccusative non copular sentences, however, can never yield a specificational sentence due to the lexical character of the predicate head.In sum, copular SSs are a kind of a byproduct which takes advantage of two independently motivated factors: the availability of an externally merged DP to the Spec of TP, as is the case with expletives to satisfy the EPP feature of T, and the non lexical nature of the Pred 0 , which is common to all kinds of copular sentences.It is thanks to the conspiracy of these two factors that a quantificational reading, namely the specificational one, can arise.When so conceived, SSs show that the computational system of human language can yield sentences which, although being bare quantificational structures devoid of any predicate-argument content, converge at the Conceptual-Intentional interface.
This interesting result is even nicer when it is taken into account that the rationale behind this approach is indeed very restrictive in that it assumes that copular sentences are as parallel to non copular sentences as possible.Let us see why.While, regarding its theta layer, a copular specificational sentence, not having an external argument in its PredP, would pattern with a non copular unaccusative one, a copular predicational sentence, by virtue of its external argument, would correspond to a non copular transitive sentence.Furthermore, this approach paves the way to analyze presentational sentences of the sort It's John (or French C'est Jean, or Romance Null Subject Language versions of it without the initial pronominal element) in essentially the same way that SSs.Presentational sentences, however, due to the expletive nature of the initial pronominal element they present, could never yield the quantificational reading associated with SSs, since they do not provide an initial DP able to be interpreted as a variable for which the postcopular DP is the value.It is not the aim of this paper to pursue this line of inquiry, but I mention it as a new avenue my proposal invites to explore.

A feature specification algorithm
An explanatory proposal for SSs requires more than a suitable structure for them, however.For the sake of concreteness, such a proposal should be able to predict why (32) below is ambiguous whereas (33) and (34) are not.The fact that this aspect is widely ignored in the literature does not free us from considering which conditions, beyond the structural ones, a copular sentence must satisfy to yield a specificational reading.In a first approximation, the following algorithm would be able to some extent to predict for a given combination of DPs around a copula whether the result will be specificational, predicational or ambiguous: (35) Feature Specification Algorithm The guiding principle: The post-copular DP in a specificational sentence can not be less specified than the pre-copular DP, where among other instantiations we universally find that (i) Singular is less specified than plural; a collective noun counts as a plural.(ii) [-animate] nouns are underspecified for person while [+animate/+human] ones are specified for person, i.e. [+3 person].(iii) First and second person personal pronouns are more specified than [+3 person] nouns or pronouns.(iv) Person is higher in the feature hierarchy than number so that when person is present in the precopular DP, number does not enter the specification competition and the postcopular DP must be congruous with the precopular one in number.
And for Romance NSL: (v) Masculine and feminine forms of articles and demonstratives, when not followed by a [-person] noun, have an optional [+person] feature and (vi) lo, in some varieties of Catalan and Spanish, and això, in some varieties of Catalan, are a [-animate, ±gender] common gender determiner; they are maximally unspecified.
Note that (32) allows for both a predicational and a specificational reading since its two DPs are singular and, then, do not contravene the guiding principle in (35).Correspondingly, (33) can only be ascriptive because of the singular in the postcopular DP vis-à-vis the plural in the precopular one.By contrast, (34) with a postcopular DP more specified than the precopular DP is specificational.
The following contrast will also fall in place by virtue of ( 35iv At first sight, (39) could be seen as a violation of the guiding principle in (35).However, a more careful observation seems to indicate that what is at stake here is a misplacement of the restrictor that instead of occuring together with the operator and the variable occurs with the value.Notice that Es cotxo més guapo és aquest ( the MSG most beautiful car is this MSG : The most beautiful car is this one) is a wellformed SS.Interestingly, the ungrammaticality of (39) and its explanation seem extendable to English: cf.*The basic one is this problem vs.The basic problem is this one.Moreover, note in this connection that there is no problem in locating the name in the postcopular DP when the sentence is predicational: This is the basic problem.
As previously advanced, the feature specification algorithm has another important application: it is at the basis of the agreement patterns found in Romance NSL.As seen supra in 2.2.5, it is not true that in these languages agreement in SSs is always with the postcopular DP.In copular sentences in general (Rosselló 2003), the copula agrees with the most specified DP according to the algorithm in (35).
One could object that this approach is incomplete, given that English type of languages show agreement with the precopular DP in all SSs and, therefore, do not follow the algorithm for agreement purposes.But I will respond to this objection saying that, first, English-like languages follow (35) to regulate the compatibility between the two DP involved in an SS and, second, that this certain degree of arbitrariness in agreement in SSs shows the spurious nature of it in this kind of sentences so that which pattern the language chooses is to a certain point irrelevant.In this sense, the wellknown fact that, at least in Romance NSL, native speakers often "make mistakes" in this kind of sentences corroborates the idea.
It goes without saying that much more work needs to be done to strengthen the validity of the proposal put forward in this paper.I expect, however, to have shown that the avenue it opens merits to be further explored.

( 16 )
El problema sou vosaltres.the problem are PRES2PL you 2PL 'The problem is you.' (17) The problem is me.
(xi)Justify the existence of prosodically neutral wellformed sequences of the form Copula DP 1 DP 2 .Certainly, in Catalan and other Null Subject Languages (NSL) one can find clauses with this order, which would constitute a counterexample for Moro's account in which the break-up of the initial symmetry acts as a trigger for the derivation to go on.Two examples of this would be the following ones: the tallest player on team.DEF is she not 'She isn't the tallest player on the team.'Thereare some further Danish data pointing at the need for a different analysis for specificational and predicate topicalization sentences.Given the impossibility, due to space limits, of illustrating them with the corresponding examples, I summarize all of them, the new ones and the preceding ones, in (11): b. Den hjeste spiller pa holdet er ikke hende.the tallest player on team.DEF is not her 'The tallest player on the team isn't her.' (10) a. DP er DP nom ikke  Predicational (predicate topicalization) b.Den hjeste spiller pa holdet er hun ikke.CP Den hjeste spiller pa holdet i SG POSS.SG 1PL son / the M.SG Jordiet not is no child 'Our son is no longer a child.'