The Focus-Background Articulation in Spanish Qualitative Binominal NPs

The goal of this article is to show that, unlike recent analyses of the qualitative binominal NP construction (QBNP) that idiot of a doctor/el idiota del médico, Spanish QBNPs require an analysis with the following features: (i) the definite article preceding the subject nominal is a fully-fledged determiner, (ii) the predicate-subject final word order is the consequence of predicate focus fronting, (iii) the subject receives background interpretation, and (iv) as a consequence of its interpretation as background, the subject of the construction must be specific, which explains the apparently syntactic restrictions on the size of the nominals occurring in this position. This proposal is shown to extend straightforwardly to a kindred Spanish nominalizing construction, yielding a better understanding of the role of focus in DP structure. In this article I consider a pretty well-studied construction: comparative qualitative binominal noun phrases (QBNPs) or N of a N construction (the most comprehensive approach is probably den Dikken 2006, whose terminology I adopt, but the bibliography is huge: Milner 1978, Ruwet 1982, Napoli 1989, Español-Echevarría 1997, Bennis et al. 1998, den Dikken 1998, Suñer 1999, Hulk & Tellier 2000, Doetjes & Rooryck 2003, Villalba 2007b). Consider a few examples (M=masculine, F=feminine, PL=plural): (1) a. el tonto del vecino the.M dumb.M of-the.M neighbor.M ‘that fool of a neighbor’ (Spanish; Suñer 1999: ex. (90a)). b. quei fessi dei tuoi fratelli those dumbos of-the.PL your.PL brothers ‘those dumbo brothers of yours’ (Italian; Napoli 1989: ex. (3)-(15)).

In this article I consider a pretty well-studied construction: comparative qualitative binominal noun phrases (QBNPs) or N of a N construction (the most comprehensive approach is probably den Dikken 2006, whose terminology I adopt, but the bibliography is huge : Milner 1978, Ruwet 1982, Napoli 1989, Español-Echevarría 1997, Bennis et al. 1998, den Dikken 1998, Suñer 1999, Hulk & Tellier 2000, Doetjes & Rooryck 2003, Villalba 2007b).Consider a few examples (M=masculine, F=feminine, PL=plural): (1) a. el tonto del vecino the.M dumb.M of-the.M neighbor.M 'that fool of a neighbor' (Spanish; Suñer 1999: ex.(90a)).b. quei fessi dei tuoi fratelli those dumbos of-the.PL your.PL brothers 'those dumbo brothers of yours' (Italian; Napoli 1989: ex. ( 3)-( 15)).c. la malparida de la teva cunyada the.F son.of.a.bitch.F of the.F your.F sister.in.law 'that damned sister-in-law of yours' (Catalan) On purely descriptive grounds, the main features of the construction can be summarized as follows: • a subject-predication structure is involved: in (1a) the nominal tonto 'fool' is predicated of the referential nominal vecino 'neighbor'; • the predicate precedes the subject: in (1a) the predicate nominal tonto 'fool' precedes the subject nominal vecino 'neighbor'; • the predicate and the subject are separated by means of preposition de 'of'; • the subject noun is preceded by an article, an indefinite one in Germanic (1a-b), and a definite one in Romance (1c-d).
The latter property minimally distinguishes the Spanish, Italian, and Catalan versions of the construction at issue from a kindred construction (attributive qualitative binominal noun phrases in den Dikken's 2006 terms): (2) a. esa mierda de libro that.F shit.F of book.M 'that shitty book' (Spanish; Casillas 2003: ex.(1a)) b. una peste di bambino a wretch of boy 'a wretch of a boy' (Italian; Napoli 1989: ex. ( 3)-( 15)) c. ton phénomène de fille your.M phenomenon.M of girl.F 'that character of a daughter of yours' (French; Hulk & Tellier 2000) Formally, this attributive variant requires the second nominal to be bare, and allows number and gender disagreement between the subject and the predicate, in sharp contrast, to the QBNPs variant in (1): (3) a. *el tonto de vecino the.M dumb.M of neighbor.M 'that fool of a neighbor' (Spanish) b. *el tonto de vecina(s) the.M dumb.M of neighbor.F(PL) 'that fool of a neighbor' (Spanish) Since this attributive variant will not be considered in this article, I will use the label QBNPs thoroughly to refer to the comparative QBNPs in (1), for the sake of simplicity (the reader is referred to Doetjes & Rooryck 1999, Suñer 1999, and den Dikken 2006 for a detailed comparison of the two constructions).
Once, we have delimited the empirical coverage of the paper, we can proceed stating the main goal of this article, which is to show that, unlike recent analyses of the qualitative binominal NP construction (QBNP) el idiota del médico 'that idiot of a doctor' (e.g.den Dikken 2006), Spanish QBNPs in particular, and most Romance ones in general, require an analysis with the following features: 1. the definite article preceding the subject nominal must be analyzed as a fully-fledged determiner, 2. the predicate-subject final word order is the consequence of predicate focus fronting (i.e.A-bar movement), 3. the subject receives background interpretation; and 4. as a consequence of its interpretation as background, the subject of the construction must be specific, which explains the apparently syntactic restrictions on the size of the NPs occurring in this position.
This analysis will be shown to account in a simple and elegant way for a variegated set of data, and to extend straightforwardly to a related nominal quantificational construction, the Spanish lo-de construction (see Bartra-Kaufmann & Villalba 2006;Villalba & Bartra-Kaufmann 2008).Yet, since it has crucial differences with respect to the standard view concerning this construction, the first section will be devoted to presenting the influential proposal developed in den Dikken (2006) for Germanic and Romance QBNPs.The second section will show the main empirical challenges that Spanish poses to such a proposal.The third section will offer a different solution to the Spanish puzzle taking into account the special information packaging of this construction, which will be argued to involve a completely different pattern from the one argued for Germanic QBNPs by den Dikken (2006); furthermore, it will add evidence supporting the analysis from the kindred Spanish lo-de nominalization construction.The last section will address the main conclusions of the paper.

The spurious article
The departing point of den Dikken (2006) account of Dutch QBNPs like die idiot van een doktor 'that idiot of a doctor' is the predication structure Following seminal ideas by Richard Kayne (Kayne 1983(Kayne , 1994) ) further developed by himself in previous work (Bennis et al. 1998, den Dikken 1998), he articulates the subject-predicate relation by means of a functional category that he calls Relator.Crucially, the head of this lexically vacuous projection may get filled by a series of elements, in the case at hand, the indefinite article, as in (5) die idiot van een doctor that idiot of a doctor Developing the work in Bennis et al. (1998), the main evidence den Dikken offers for considering the indefinite article een 'a' spurious is number agreement.In sharp contrast with English, which only allows singular nouns after the indefinite article (*those idiots of a doctors), Dutch does allow plural nouns as well, and, crucially, the singular article een 'a' must be inserted regardless of the plural number of the following noun (Bennis et al. 1998, exs. (12a), (12b), ( 16) and (17b), respectively): (6) a. dat tuig van een voetbalsupporters that scum of a soccer.supportersb. die schatten van een kinderen those darlings of a children This contrasts with the behavior of the real indefinite article (example adapted from Bennis et al., 1998, ex. (11)): (7) Ik heb een boek/*boeken gelezen.I have a book/books read His conclusion is that the indefinite article in Germanic QBNPs is just filling the slot for the Relator head, and should be considered spurious.

The size of nominals
Since the indefinite article is the head of the Relator head, and, consequently, doesn't form a constituent with the subject of QBNPs, one must determine the size of this nominal constituent.According to den Dikken (2006), the subject and the predicatemust be smaller than a DP, particularly Number Phrases (I will refer to this proposal as the NumP Hypothesis).He grounds this proposal on two pieces of evidence.On the one hand, the subject nominal may show number disagreement with respect to the whole DP, which is the one that agrees with the verb when required: (8) die ramp van een feiten ?komt/*komen zeer ongelegen those disaster of a facts comes/*come very inconvenient The Focus-Background Articulation in Spanish Qualitative Binominal NPs 135 This behavior leads den Dikken to argue that the subject of QBNPs must be big enough to contain a NumP, independent of that of the whole DP.On the other hand, the subject of QBNPs cannot host material arguably located in a high position within the (extended projection of the) DP, above NumP, namely quantifier phrases, as the ungrammaticality of the following Dutch example suggests: (9) *die ramp(en) van (een) alle feiten / ieder feit those disaster(s) of a all facts / every fact This restriction, den Dikken claims, shows that the subject must be clearly smaller than a full DP, and together with the evidence regarding agreement, he concludes that the subject must be at least and at most a Number Phrase.

Predicate Inversion
The third important feature of den Dikken's analysis involves Predicate Inversion.Following the analysis of inverted copular sentences in Moro (1997), Bennis et al. (1998)  Den Dikken extends this line of research to QBNPs, so that the Relator head incorporates to the head of a higher functional phrasethe Linkerto enable further raising of the predicate:

Information structure
Finally, den Dikken links Predicate Inversion with information structure.Following a suggestion in Lagae (1994), den Dikken & Singhapreecha (2004: 8) claim that "[w]hen one inverts a predicate around its subject, the result is an information-structure representation in which the postcopular noun phrase is invariably the focus".So then, they argue for an information packaging of QBNPs like the following: So then, we can summarize the account of QBNPs by den Dikken (2006) in three main points: 1. the existence of a spurious (in)definite article, 2. the size restriction on the subject to NumPthe NumP Hypothesis -, and 3. the existence of Predicate Inversionwhich yields a specific information packaging, where the predicate is old information, and the subject new information.
In the following section, we will review the challenges Romance QBNPs pose to den Dikken's analysis.
The Focus-Background Articulation in Spanish Qualitative Binominal NPs 137 2. From Germanic to Romance: Problems for the NumP hypothesis

Cross-Romance variation
The picture one gets after the reading of den Dikken ( 2006) is that a clear-cut parameterization exists separating Germanic and Romance QBNPs: whereas Germanic allows a spurious indefinite article, Romance resorts to a spurious definite article.Yet, the presumed parametric difference stemming from this picture is both too restrictive, for the subject in QBNPs can be headed by an indefinite article under certain conditionsa fact originally pointed out by Napoli (1989, 203) for Italian, and acknowledged in a footnote by den Dikken (2006, 297 fn.54): (15) a. quel fesso di un vicino that dumbo of a neighbor (Italian; Napoli 1989: (4)-( 23)) b. la idiota de una vecina amiga mía the.F idiot.F of a.F neighbor.F friend of.mineHence, to extend the analysis of Germanic QBNPs to Romance is both too strong a proposal, as far as the presumed spurious article is concerned.

Agreement
Den Dikken's claim that the definite article in Romance QBNPs is spurious in the same way the indefinite article in Germanic QBNPs runs into trouble when agreement is concerned.We have seen that the main evidence for the spurious nature of the indefinite article een 'a' in Dutch QBNPs was the lack of agreement with the subject nominal (I repeat example (6a) for the sake of reference): (17) a. los idiotas de los vecinos the.PL idiots of the.PL neighbors b. la idiota de una vecina amiga mía the.F idiot.F of a.F neighbor.F friend of.mineSo then, no empirical evidence exists whatsoever to assume that the article is spurious in Spanish QBNPs.

The NumP Hypothesis
Another major point of den Dikken's analysis that doesn't prove successful when extended to Romance QBNPs is the NumP Hypothesis, which states that both the subject and the predicate nominal are at least and at most NumPs.We are not discussing the case of the predicate nominal, but strong evidence exists against such a size restriction regarding the subject of Romance QBNPs.First, certain quantifiers (see 2.4 below) are possible within the subject in Spanish QBNPs, even though one usually places them above the NumP, which is den Dikken assumption: (18) los idiotas de algunos/muchos (de los) alumnos the.M.PL idiots of some.M.PL/many.M.PL of the.M.PL students Another problem, which den Dikken is well aware of (see den Dikken, 2006, fn. 63), concerns the presence of demonstratives within the subject of QBNPs, for they are typically assumed to occupy a position in the DP higher than NumP (see Giusti 1993, andRoca 1997 Therefore, it is clear that the behavior of quantifiers and demonstratives doesn't support the extension of the NumP Hypothesis to Romance QBNPs.

Information structure in Romance QBNPs
As we have seen in 1.4, the information packaging associated with Germanic QBNPs conforms to the one expected for a Predicate Inversion structure, namely the predicate is interpreted as old/given information and the subject is interpreted as new information.However, such an information packaging doesn't fit in with Romance QBNPs.Let us see, in the first place, that the subject position cannot be a focus position through two standard tests for focushood.First, the association with focus particles like only is usually a reliable test to determine the focus-background material within a constituent: (20) a.Since right-dislocates are background information, we can make the prediction that the subject of QBNPs in Spanish can enter into backward pronominalization.This prediction is borne out: (26) Su i procesamiento deprimió al corrupto del alcalde i .his prosecution depressed to-the corrupt of-the mayor 'His i prosecution depressed the corrupt mayor i .' Another strong piece of evidence comes from the contrast between inherently nonspecific and specific quantifiers within the subject position.As acknowledged in 2.3, quantifiers were admitted in the subject nominal, against the predictions of the NumP Hypothesis.Yet, the interesting point is that only specific quantifiers are possible in this position.Consider the contrast between Spanish nonspecific todo 'every', cualquier 'any', and demasiados 'too many' with inherently specific quantifiers todos los 'all of the' and ambos 'both': (27) a. *Conoció al idiota de todo/cualquier alcalde met to-the idiot of every/any mayor b. *Conoció a los idiotas de demasiados/excesivos met to the.M.PL idiots of too.many/too.manyalcaldes mayors (28) Conoció a los idiotas de {todos los/ambos} alcaldes met to the.M.PL idiots of all the.M.PL/each mayors We will turn back to these data and further evidence in 3.2, but from the small set of data just presented, the conclusion is clear that the subject position of QBNPs in Romance is better analyzed as background, which suggests that the information packaging of QBNPs is exactly the opposite to the one defended for Germanic QBNPs by den Dikken & Singhapreecha (2004)

A new proposal for Romance QBNPs
In the previous section we have seen that the extension of den Dikken (2006) analysis of Germanic QBNPs to Romance is at least problematic.We have seen that there is no evidence for a definite spurious article in Romance QBNPs, nor for analyzing the subject of QBNPs as a NumP.Furthermore, we have concluded that the information structure of QBNPs in Romance doesn't fit the typical schema associated with Predicate Inversion in Germanic.The next step is to build up a new proposal for Romance QBNPs grounded on the different information structure involved.
We have just reviewed in 2.4 evidence that the information partition of Romance QBNPs is the opposite of the one proposed by den Dikken & Singhapreecha (2004), den Dikken (2006).Namely, • the subject of Romance QBNPs is a background topic; • the nominal predicate of Romance QBNPs is focus.
These two features will become paramount in accounting for the complex and intricate set of data reviewed in the previous section.Let us spell out the analysis in detail.

The analysis
We will take as a point of departure the Relator Phrase that articulates the basic subject-predicate structure, as proposed by den Dikken (1998Dikken ( , 2006) ) and den Dikken & Singhapreecha ( 2004): Here agreement between the subject and the predicate holds without any intervening element, yielding the full range of realizations that we find in Romance (see 2.2).
The next step in the derivation involves merge of Focus, which probes and attracts the predicate NP (on the presence of a FocusP in the DP, see Aboh 2004, Giusti 1996, Demonte 2008, Villalba & Bartra-Kaufmann 2008): Two points are worth commenting here.First, as noted by one anonymous reviewer, we should find some independent evidence supporting the analysis of de as a focus head/marker in this structure.One piece of compelling evidence is the following Spanish degree construction (see Gutiérrez-Rexach 1999, Villalba 2003 As the answer-question pair reveals, the focus is placed on the indexical degree adverb así 'so', and the presence of the focus marker de 'of' is obligatory.This is exactly what happens with the exclamative version, where a null degree operator seems at stake, forcing the presence of the marker, as well: (33) ¡Son de fuertes!are.3PL of strong 'They are so strong!' The second aspect of the proposed structure in (31) that merits a comment is the nature of the movement involved.Crucially, this is a standard case of predicate non-A-movement, and not one of Predicate Inversion in Moro (1997), Bennis et al. (1998), andden Dikken (2006) terms.Yet, this departure from the standard analysis of QBNPs seems accurate on empirical and theoretical grounds.On the one hand, it parallels cases of predicate focalization, which are instances of non-A movement, and offers us a principled explanation of the fact that QBNPs are islands for extraction: the NP in Spec,FocP counts as an A'-intervener for further A'-movement.
(34) *¿[De qué pueblo] conoció al idiota del alcalde t? of which village met to.the idiot of.themayor On the other hand, another direct payoff of the focus analysis concerns information structure.It creates a direct mapping for the focus-background articulation in exactly the same way FocusP does in the system by Rizzi (1997): Finally, note that the particle de 'of' and the subject nominal do not form an XP constituent, which is at the basis of the ungrammaticality of the following Spanish examples: The final step involves merge of D, which will probe and match the uninterpretable phi-features of the predicate NP, surfacing as an agreeing determiner: The Focus-Background Articulation in Spanish Qualitative Binominal NPs 143 After showing the technical part of the analysis, now the time is ripe to explain how the size restrictions apparently derived from den Dikken's (2006) NumP Hypothesis follow from the proposal in the text without additional stipulations.

Testing the analysis: Referentiality restrictions
As we have seen in 2.3 and 2.4, Spanish QBNPs allow the presence of quantifiers in the subject position, against the predictions of den Dikken's (2006) NumP Hypothesis.Yet, we pointed out that not all quantifiers were allowed.In this subsection, we consider the issue in some detail, for it brings us compelling evidence for supporting the information packaging that we propose for Spanish QBNPs.
Even though complex and intricate, the distribution of quantifiers within the subject nominal shows a clear underlying pattern once their semantic contribution is considered.The key point is specificity: the more specific the quantifier, the better its occurrence in the subject position.Consider again the case of universal quantifiers: (38) a. Conoció a los idiotas de {todos los/ambos} alcaldes.knew to the.M.PL idiots of all the.M.PL/each mayors b. *Conoció al idiota de todo/cualquier alcalde.knew to-the.M idiot of every/any mayor Inherently specific quantifiers (38a) are fine, whereas inherently nonspecific ones (38b) are impossible, in the subject position.Crucially, if we force a specific interpretation by means of a partitive structure, the phrase with the quantifier cualquier 'any' improves: (39) ?Conoció al idiota de cualquiera de aquellos alcaldes.knew to-the.M idiot of any of those mayors Since partitivity is standardly tied to specificity (see Enç 1991), we predict that the possibility of obtaining a partitive reading will be an important factor in the distribution of quantifiers in the subject position of QBNPs.The data confirm this prediction in full.Let us see a complex but illustrative case.Whereas monotone increasing quantifiers typically allow a partitive reading, monotone decreasing and nonmonotone quantifiers don't.As a consequence, the former are fine in the subject position of QBNPs: ( From the set of data just reviewed, it seems fair to conclude that the subject of QBNPs in Romance must be specific, which is precisely the expectation, given the proposal that this position must get a background information interpretation. Yet, this fact fits in with the information structure evidence presented in 2.4, for we know from backgrounding strategies like right-dislocation (RD) that specificity is a requisite for becoming a background topic (see Villalba 2000: ch. 3).Consider the behavior of Catalan right-dislocation with respect to the referentiality restriction just described (the language choice is purely instrumental, for Catalan allows RD more easily than Spanish does; see Villalba 2007a).On the one hand, inherently specific quantifiers and quantifiers allowing a partitive reading can be right-dislocated: (43) a.On the basis of this parallel behavior, we can conclude that the apparent specificity restriction on the subject position of QBNPs is an epiphenomenon deriving from the underlying background interpretation assigned to such a position, which follows straightforwardly from the analysis put forward in this section.

Extending the analysis: The Spanish lo-de construction
In this section, I extend the information-based approach developed in 3 to an independent nominal construction: the Spanish "lo-de construction" (see Gutiérrez-Rexach 1999, Bartra-Kaufmann & Villalba 2006  On the basis of these data, we can conclude that, as far as their basic features are considered, the Spanish lo-de construction can be analyzed on a pair with QBNPs, along the lines presented in 3.1.Now, it is time to turn our attention to the referentiality restrictions affecting the subject in this Spanish construction, which will be shown to parallel the ones found in Romance QBNPs generally. As observed by Bartra-Kaufmann & Villalba (2006), there is a referentiality constraint affecting the subject of the lo-de construction.Consider the case of nonspecific indefinite DPs (48a), bare plurals (48b), and NPI and downward entailing quantifiers ( 48c Bartra-Kaufmann & Villalba (2006) do not offer an explanation for this behavior, it obviously parallels the facts discussed in 3.2.The next step is determine whether it makes sense extending to the lo-de construction the analysis proposed in 3.1 for QBNPs.The answer seems positive, for the information packaging of the lo-de construction is identical to that of QBNPs, namely the fronted predicate is focus and the subject is interpreted as part of the background.
On the one hand, the subject DP cannot be focus, since it cannot be associated with the focus particle sólo 'only' nor be a wh-element in situ (compare with the QBNPs exs.( 21) and ( 23 On the other hand, backward pronominalization provides us with evidence that the subject of the lo-de construction is background information (for QBNPs, see (26) above): (50) Su i autor se sorprendió de lo exitoso d[el libro] i .his/her author self surprised of LO successful of-the book These data confirm our previous description of the QBNPs data: the predicate is interpreted as focus, whereas the subject is interpreted as background information.This particular information packaging gives us a principled and independently motivated understanding of the quantificational restriction applying in the lo-de construction, which in turns strongly confirms the analysis of Spanish QBNPs developed in 3.1.

Conclusions
In this paper, it has been argued that Spanish QBNPs don't fit the analysis advocated for by den Dikken (2006), namely that there is no definite spurious article in Romance QBNPs, that the subject of QBNPs is not a NumP in Romance, and that the construction involves A-bar predicate movement to a DP-internal Focus Phrase, which renders the subject of Spanish QBNPs a background topic.This analysis has been shown to succeed in explaining the complex and intricate pattern of referentiality restrictions in the subject position of QBNPs in a simple and straightforward way.Finally, the analysis defended has been successfully extended to a different Spanish nominalizing construction: the lo-de construction.
It will be interesting to extend this analysis to Romance QBNPs in general, which is beyond the limits of this article.For the time being, there is ample evidence that Italian and Catalan QBNPs behave similarly to Spanish ones (see Napoli 1989 for Italian, and Villalba 2007b, for Catalan).Yet, as pointed out to me by Jean-Yves Pollock, and one anonymous reviewer, French seems to offer a totally different picture, closer to the Germanic type (something similar can be said of Romanian, according to the data in Visan 2003).Yet, until a finer-grained analysis of Romance QBNPs as a whole is offered, I hope that, even though limited in coverage, the data and the proposal presented in this article will pave the way for a better understanding of this amazingly complex nominal construction.

(
12) [that idiot] OLD of [a doctor] NEW 1.5.Romance QBNPs As we have shown before, Romance QBNPs must substitute the indefinite article by a definite one before the subject nominal (I repeat the examples in (1), adding italics to the relevant definite articles): son.of.a.bitch.F of the.F your.F sister.in.law 'that damned sister-in-law of yours' (Catalan) As a consequence, den Dikken just applies the structure and analysis just reviewed, substituting the indefinite article by the definite one in the head of the Relator Phrase: (14) a. quello ignorante del doctore that ignoramus of-the doctor 'that ignoramus of a doctor' (Italian; den Dikken 2006) b In contrast, the (in)definite article in the subject of Romance QBNPs agrees in gender and number, as the following Spanish examples show: idiot b. *¿[De quién] te presentaron al tonto t? of who to.you introduce to-the.M dumb.M Mary had [a lamb] focus only.b.Only [Mary] focus had a lamb.When extended to QBNPs, the conclusions are clear: the focus particle sólo 'only' cannot associate with the subject.Consider the following Spanish examples: (21) a. *¿Recuerdas al burro de sólo aquel médico?remind.youto-thedonkey of only that doctor b. *Hablé con el granuja de sólo aquel alcalde.talked.I with the crook of only that mayor A second test concerns wh-in situ.When in final position, Spanish whelements are focus, and must precede any right-dislocate (see Etxepare &Once we have shown that the subject of Spanish QBNPs cannot be focus, let us see now that the subject is typically interpreted as background.One piece of evidence stems from backward pronominalization, which is much more restricted a phenomenon in Spanish than in English: Bartra-Kaufmann & Villalba (2006)ive.F of the house.F 'It surprised me how expensive the house was.' Syntactically, the lo-de construction is headed by the so-called 'neuter article' lo, which combines with a gradable nonagreeing adjective, which in turn establishes a predication relation with a DP, but shows no agreement, but rather a default third person singular form.AsBartra-Kaufmann & Villalba (2006)observe, constituency tests, like wh-movement, fail for the apparent PP headed by de 'of', just as happens with QBNPs (see 3.1): Villalba take this behavior to indicate that de is not a true P in these constructions, but rather a Linker, on a parallel with QBNPs.Finally, there is another basic property that makes the lo-de construction and QBNPs similar: islandhood.As the following examples illustrate, the subject position of former is an island for extraction (for QBNPs, see 3.1 above): to