Hostname: page-component-74656f9884-8r8nv Total loading time: 0 Render date: 2026-11-07T03:49:16.960Z Has data issue: false hasContentIssue false
  • English
  • Français

The Big Mess Construction: interactions between the lexicon and constructions1

Published online by Cambridge University Press:  08 June 2011

JONG-BOK KIM  and
PETER SELLS
JONG-BOK KIM
Affiliation:
School of English, Kyung Hee University, Seoul, 130-701, Koreajongbok@khu.ac.kr
PETER SELLS
Affiliation:
Department of Linguistics, SOAS, London WC1H 0XG, United Kingdomsells@soas.ac.uk
Rights & Permissions [Opens in a new window]

Abstract

The so-called Big Mess Construction (BMC) (e.g. so prominent a punctuation), introduced by a limited set of degree words, places an adjectival expression in the predeterminer position. In movement approaches, such idiosyncratic properties of the BMC have been attributed to the interaction of functional projections and movement operations, whereas in surface-oriented analyses focus has been placed on the supposition of special constructions and their constructional properties. In this article, we show that neither of these two previous perspectives captures the variations and flexibility of the construction in question satisfactorily. Our approach adopts the view that degree words are functors selecting their head, and attributes the peculiarities to the interactions between the lexical properties of the degree items and the constructional constraints in question.

Information

Type
Research Article
Information
English Language & Linguistics , Volume 15 , Issue 2: The structure of the noun phrase in English: synchronic and diachronic explorations , July 2011 , pp. 335 - 362
Copyright
Copyright © Cambridge University Press 2011

1 Introduction

The so-called Big Mess Construction (BMC)Footnote 2 exemplified by corpus examples like (1) is peculiar in that it has predeterminer adjectival elements followed by an NP with the indefinite article:Footnote 3

  1. (1) so-type

    1. (a) Hunger was now [so powerful] [a force] in its life. (BNC HTM W_ fict_prose)

    2. (b) Companies with [as strong] [a balance] sheet as yours have been known to seek acquisitions. (BNC HYE S_meeting)

    3. (c) It tells us just [how big] [a mess] the Government has got us into. (BNC K5M W_newsp_other_report)

    4. (d) We have far [too great] [a gap] between these two states. (BNC KRG S_brdcast_discussn)

    5. (e) [This new] [a phoneme] would have two allophones. (BNC K93 W_ac_soc_science)

    6. (f) It's about [that big] [a diameter]. (BNC KCS S_conv)

    7. (g) He proved far [more successful] [a dealer] than he had a client. (BNC EUU W_commerce)

As illustrated here, this BMC can be introduced by a degree word, but only one of a limited set of degree words including so, as, too, how, this and that licenses the construction. For example, degree expressions like somewhat or very do not allow the BMC (cf. Bolinger Reference Bolinger1972; Bresnan Reference Bresnan1973; Huddleston & Pullum et al. Reference Huddleston and Pullum2002; among others):

  1. (2)

    1. (a) He has [a somewhat different view] (*[somewhat different a view]) of how strong the economy is coming back now. (COCA SPOK Fox_ Sunday)

    2. (b) I have grown up with movies as [a very important part] (*[very important a part]) of my life. (COCA NEWS USAToday)

In addition to this so-type given in (1), English also allows a similar BMC-style construction, as attested by naturally occurring data:

  1. (3) such-type

    1. (a) Identity is [such] [a powerful force] because it opens a world of meaning larger than physical and material life. (COCA MAG USNWR)

    2. (b) At that time, I didn't realize [what] [a big deal] that was. (COCA NEWS Houston)

    3. (c) They'd been rode [many] [a long mile]. (COCA FIC New Yorker)

The examples show us that like the so-type BMC, the adjectival such, exclamative what, and quantifier many can also appear in the NP-external position.Footnote 4

These two types, both consisting of adjectival predeterminers (APD) and an indefinite head NP, display several similarities. Both the so-type and such-type require the NP to have the indefinite article a/an. The similarities can also be observed in the very similar meanings:Footnote 5

  1. (4)

    1. (a) He failed to grasp why such a small thing unsettled her. (COCA MAG Ms)

    2. (b) He failed to grasp why so small a thing unsettled her.

In both examples here, there is a small thing that unsettled her: the only difference lies in the nuances of so and such.

Another intriguing property of these two types arises when some of these APD expressions are combined with a discontinuous modifier (DCM):

  1. (5)

    1. (a) It was [so big] a city [that the bishop and the premier of Quebec were living there]. (COCA FIC CanadianAGeograpic)

    2. (b) Surely this wasn't [too high] a price [to pay for all my worldly success]. (COCA FIC FantasySciFi)

    3. (c) It will be much [more widespread] a solution [than just looking at those people]. (COCA SPOK NPR_Science)

    4. (d) Optical systems don't do [as good] a job [as their electronic counterparts]. (COCA NEWS AssocPress)

  2. (6)

    1. (a) He does [such] a good job [that it hardly seems worthwhile to compete]. (COCA MAG MotherEarth)

    2. (b) You don't have half [such] a hard time [as I do]. (COCA FIC Bk: LittleWomen)

In both (5) and (6) here, there is a syntactic dependency between the APD elements and a second bracketed element such as a CP, an infinitival VP, a than-comparative clause, or as-relative clause. The dependency between the APD and DCM is tightly constrained in syntactic aspects, as observed from the following:Footnote 6

  1. (7)

    1. (a) We found [so] dark a cave that/*why we could not see a thing.

    2. (b) You don't have half [such] a hard time as/*like I do.

The APD so dark or such requires an indefinite NP, but also places a restriction on the possible type of its DCM: each needs to be introduced by that and as, respectively. No other but similar expressions can introduce the DCM.

As we have seen, both so-type and such-type display nontrivial complexities in English syntax. In dealing with these two empirically and theoretically challenging types of BMCs, there have been two main perspectives: movement approaches (e.g. Kennedy & Merchant Reference Kennedy and Merchant2000; Vikner Reference Vikner, Adamzik and Christen2001; Matushansky Reference Matushansky2002; Wood Reference Wood2002; Wood & Vikner Reference Wood, Vikner, Sleeman and Perridon2011) and construction-based approaches (e.g. Van Eynde Reference Van Eynde and Müller2007; Kay & Sag Reference Kay and Sag2009). In the former, the APD expression is traditionally moved out of the source NP structure, whereas the latter generates the APD and the BMC with no movement operations but by constructional constraints. As we will see in due course, each of these two perspectives encounter theoretical or empirical issues. The treatment of the DCM construction has also been a challenge to both movement and non-movement syntactic analyses. Within P&P or Minimalist approaches, issues arise as to what motivates the necessary movements (cf. González Escribano Reference González Escribano2005). Within nonderivational analyses (cf. de Mönnink Reference de Mönnink and Kirk2000; Kay & Sag Reference Kay and Sag2009), difficulties arise as to how to differentiate the dependency of the DCM, which is not truly ‘long-distance’, from canonical unbounded dependency constructions such as wh-constructions.

In this article, we first review some basic properties of these APD and DCM constructions on the basis of the previous literature and a corpus search (mainly using the ICE-GB, COCA and BNC). We conclude that none of the previous approaches accounts satisfactorily for the variations and flexibility of the two constructions. Drawing on corpus findings, we attempt to provide a lexicalist analysis of these two constructions that places the nature of these intriguing constructions in the interactions between lexical properties and constructional constraints, rather than in movement processes or constructions alone.

2 Properties of the adjectival predeterminers

As we have seen, the so-type and such-type APDs are similar in that they both can appear in a predeterminer position, but differ with respect to the position of AP: only the so-type allows an AP to be in the predeterminer position, as shown in the following contrast (cf. Abney Reference Abney1987; Huddleston & Pullum et al. Reference Huddleston and Pullum2002):

  1. (8)

    1. (a) It had been almost two years, and she had been [such] a [young] child. (COCA FIC Storyworks)

    2. (b) *It had been almost two years, and she had been [such young] a child.

  2. (9)

    1. (a) So many things could happen to [so young] a child. (COCA FIC Bk:WindDancerStorm)

    2. (b) *So many things could happen to [so] a [young] child.

As noted, both such- and so-type APDs combine only with the indefinite article a/an: no other determiner is licensed (cf. Van Eynde Reference Van Eynde2006):Footnote 7

  1. (10)

    1. (a) *How serious some problem is it?

    2. (b) *They are so good Ø bargains I can't resist buying them.

  2. (11)

    1. (a) *The drug was such any serious problem that there is no other way to deal with it.

    2. (b) *I didn't realize what many big deals they are.

With respect to the use of indefinite article, unlike the so-type, such does not always require an indefinite article. It can combine with a bare NP or plural NP (cf. Siegel Reference Siegel1994; Wood Reference Wood2002; Spinillo Reference Spinillo2003):

  1. (12)

    1. (a) Thank you for coming on [such] [short notice]. (COCA SPOK CBS_sixty)

    2. (b) He never mentioned they were [such] [good friends]. (COCA FIC Bk: BordeauxBetrayal)

This cannot happen with so:Footnote 8

  1. (13)

    1. (a) I am lucky to have *so/such [good friends].

    2. (b) I am not in *so/such [good shape] after all.

The so- and such-types have different uses. The so-type expressions are degree adverbs, but the such-type expressions can be either degree or other interpretations. For example, such has a ‘kind’ reading especially when it is used in the postdeterminer position:Footnote 9

  1. (14)

    1. (a) Other European countries, such as France and Italy, were excluded because [no such uniform pattern] could be found there. (BNC CLE W_ commerce)

    2. (b) These letters, on their true construction, did not give rise to [any such implied agreement]. (BNC FCL W_ac_point_law_edu)

The expression such in both examples here is in the adjective position, denoting a ‘kind’. Meanwhile, the degree adverb so cannot be in the NP-internal position even with the determiner no or any:

  1. (15)

    1. (a) [No such/*so standard interface] is supplied reducing the available choice to a handful of units. (BNC C00 W_commerce)

    2. (b) The law did not recognise [any such/*so general principle]. (BNC FCL W_ac_polit_law_edu)

Further intriguing properties arise from the lexical idiosyncrasies of so-type expressions. Of the so-type degree words, the comparative more and less appear in either the predeterminer or the postdeterminer position:

  1. (16)

    1. (a) Any of these could be a sign of [a more serious problem] that needs medical attention. (COCA MAG GoodHouse)

    2. (b) I think it's [more a pragmatic approach], where they want to choose a candidate who can actually win. (COCA SPOK NPR_ATCW)

Our corpus findings indicate that there is no clear preference over the other ordering: we find almost equal frequency for both cases.

As for the other degree words too, that, this and enough, the previous literature has often taken it that they cannot appear in the postdeterminer position, but corpus findings indicate that they can appear in the canonical prenominal position, whose frequency we cannot ignore:

  1. (17)

    1. (a) It was always his conviction that feeling and character must take precedence over [a too literal representation] of anatomy. (COCA NEWS CSMonitor)

    2. (b) But [a that hard schooling] would avail him nothing here. (COCA FIC BkSF:DragonToken)

    3. (c) If that's not [a good enough excuse], he has others. (COCA NEWS USAToday)

A further complication arises from the possibility of having the preposition of between the APD and the head noun. The preposition of is often found in actual texts, though there is known dialectal variation (Zwicky Reference Zwicky1994; Kennedy & Merchant Reference Kennedy and Merchant2000):

  1. (18)

    1. (a) They and their parents are proud of what they have accomplished in [so short of a time]. (COCA ACAD MusicEduc)

    2. (b) Maybe I'm not [as good of a parent] as I thought I was. (COCA SPOK CBS_Sixty)

    3. (c) But no one seemed to want to stay there, and Augusta wasn't willing to let anyone create [too wide of a margin]. (COCA NEWS AssocPress)

    4. (d) The superintendent told the audience that the poor results ‘should not have been [that big of a surprise]’ because of a change in the math curriculum. (COCA NEWS AssocPress)

    5. (e) ‘As soon as you say that your mom is in the penitentiary, people wouldn't understand [how great of a lady] she really is, how caring she is,’ Warren says. (COCA NEWS USAToday)

The preposition of here seems to have no semantic content, but the restriction on the indefiniteness ensures that the preposition is a member of this BMC:

  1. (19)

    1. (a) *They are proud of what they have accomplished in [so short of the time].

    2. (b) *Maybe I'm not [as good of the parent] as I thought I was.

Corpora examples present an extra puzzle, with the doubling of the indefinite determiner with such:

  1. (20)

    1. (a) This was the first he had heard of [a such a preposterous event]. (COCA FIC FantasySciFi)

    2. (b) So far as I could see, there was [no such a thing] as a New Yorker story. (COCA ACAD AmerScholar)

These examples indicate that at least to some speakers, the combination of such with an indefinite NP does not close off the NP projection; the resulting phrase is a still Nʹ which can host a limited set of determiners.

What we have observed so far is the complex structure involving the BMC, including two different types of APD – so- and such-types. Even though the APD expressions can be classified into these two types, each degree item also displays its own peculiar properties. Their distributional possibilities are much more flexible than the previous literature has generally assumed (cf. Spinillo Reference Spinillo2003).

3 Interaction with the discontinuous modifier

As noted earlier, some of the so-type APDs are linked to a DCM element.Footnote 10 For example, the so-type APD can be linked to a finite CP, an infinitival VP, a than clause, or even a pseudo-relative clause headed by as:

  1. (21)

    1. (a) [so big] a business [that we couldn't afford to do an acquisition]

    2. (b) [too high] a price [to pay for the moon]

    3. (c) [more eloquent] a comment on the election [than any number of Sinn Fein leaflets]

    4. (d) [less important] a variable [than gender]

    5. (e) [as intelligent] a man [as I would like to be]

The APD and the dependent here are interrupted by the head NP. The dependence between APD and DCM can also be observed with such:

  1. (22)

    1. (a) Don has [such] a big house [that I actually got lost on the way to the bathroom].

    2. (b) Shelly has [such] beautiful eyes [that she got a job as a make-up model].

Note that the DCM can be introduced even when so or such is used as a non-APD expression:

  1. (23)

    1. (a) The music is [so] loud [that I can't sleep].

    2. (b) She spoke [so] quickly [that I can't understand her].

  2. (24)

    1. (a) There's no [such] thing [as a definite statement].

    2. (b) I mean there are [such] things [as political groupies].

In the examples (23), the degree items are not in the predeterminer position. Further, such in (24) is not in the APD position but still introduces a DCM expression. This implies that it is not the APD but a simple degree lexical expression so or such that may introduce the dependency.

We also note that not every degree expression is linked to a DCM (cf. Kay & Sag Reference Kay and Sag2009):

  1. (25)

    1. (a) [This] big a business (*[that we couldn't afford to do an acquisition]) is going to take a long time to get over.

    2. (b) [How] high a price (*[to pay for the moon]) will they go to?

Given these observations, we can have at least the following three different types of expression with respect to the APD (cf. Bolinger Reference Bolinger1972; Sadler & Arnold Reference Sadler and Arnold1994; Huddleston & Pullum et al. Reference Huddleston and Pullum2002; Kay & Sag Reference Kay and Sag2009):

  1. (26)

    1. (a) APD and DCM both possible: so, too, more, as, enough, such, . . .

    2. (b) APD only: this, that, how, what, . . .

    3. (c) DCM only: same. . .[as. . .], similar. . .[to. . .], Adj-er. . .[than. . .], differ. . .[from. . .], superior. . .[to. . .]

In sum, the two types of APD are different in many respects, and not even all members in the same type behave alike. For example, the degree adverb so and this are different with respect to the licensing of a DCM; such and what are also different in this sense. The observations given here reflect the fact that the lexical properties of these expressions play an important role in controlling the BMC as well as the DCM.

4 Functional projections and movement analyses

As pointed out by Kennedy & Merchant (Reference Kennedy and Merchant2000), the external syntax of the BMC has presented challenges to movement approaches since, in traditional wisdom, the elements on the left branch cannot be extracted:

  1. (27)

    1. (a) *How many does he have [__ books] in the bookcase?

    2. (b) *How interesting did Pico write [a __ novel]?

  2. (28)

    1. (a) *How much was he [__ unusual]?

    2. (b) *How did Tom wrote [an __ interesting novel]?

The ungrammaticality of these examples tells us that neither a full AP nor a degree term can be extracted out of the left-branch position of the head NP or the attributive position (Corver Reference Corver1997, Reference Corver, Coene and D'hulst2002). The BMC has exactly the same configuration in which the APD in the left branch appears to be extracted out of the NP:

  1. (29)

    1. (a) Hunger was so powerful [a __ force] in its life.

    2. (b) How big [a __ challenge] do you want?

To generate such BMC-involving sentences, P&P approaches have allowed the attributive AP to move into the Spec of DP (cf. Corver Reference Corver1990; Giorgi & Longobardi Reference Giorgi and Longobardi1991; Zamparelli Reference Zamparelli1995; Haegeman & Guéron Reference Haegeman and Guéron1999). For example, consider a simple configuration given in Kennedy & Merchant (Reference Kennedy and Merchant2000: ex. (47)):

  1. (30)

In this kind of approach, the key property is the motivation for the movement for feature checking, as set forth in detail by Kennedy & Merchant (Reference Kennedy and Merchant2000). The feature [+wh] is assigned on the DegP, and the phrasal movement is a type of wh-movement. The [+wh] feature on DegP is then passed to the head of DP for agreement reasons and the DegP moves to the Spec of DP to create an interpretable feature relationship between a [+wh] DegP and the head of the nominal constituent in which it originates (Corver Reference Corver1990; Merchant Reference Merchant1999). The feature [+wh] is used to explain the following ungrammatical example (Hendrick Reference Hendrick and Halpern1990; Moro Reference Moro1997; Lechner Reference Lechner1999; Kennedy & Merchant Reference Kennedy and Merchant2000):

  1. (31) *Pico wrote a more interesting novel than [CP Opi Brio wrote [DP tʹi a[+wh] [NP ti play]]].

In the analysis of Kennedy & Merchant (Reference Kennedy and Merchant2000), the comparative operator, functioning as the [+wh] attributive, moves to the Spec of DP and then to the Spec of CP. Spec–Head agreement ensures that this [+wh] feature is transferred to the head D. These authors propose (Reference Kennedy and Merchant2000: 112) that there is no lexical item in English which can be realised in this D position, and hence the derivation crashes at PF, as some grammatical information cannot be lexicalized.

However, this account is not all there is to say, since the APD need not include a wh-expression, as we have seen earlier:

  1. (32)

    1. (a) You can do [as good a job] or better.

    2. (b) My life was going to be [this huge a soap opera].

    3. (c) There is [no more noble a sacrifice].

    4. (d) AIDS is [bad enough a problem].

In addition, note the existence of the apparently meaningless of before the indefinite article:

  1. (33)

    1. (a) [How long of a novel] did Brio write?

    2. (b) I ate [too big of a piece].

    3. (c) Bob didn't write [as detailed of a proposal] as Sheila did.

    4. (d) If I ever see [that disgusting of a movie] again, I'll ask for my money back.

Even though the of insertion in such examples is not accepted in every dialect of English, our corpus examples do include such variation. Such examples motivated Kennedy & Merchant (Reference Kennedy and Merchant2000) and others to introduce the functional projection FP, a higher layer above the DP:

  1. (34)

In this revised configuration, the landing site of the DegP is now the Spec of FP. In particular, the head F accommodates the preposition of, capturing the ordering of APD followed by a DP.

However, note that this configuration is still not complete, when considering the complexity of the BMC. Leaving aside the issue of how to ensure the D value is not any indefinite but only a/an, the analysis still does not address the issue of what motivates the movement of the left-branch elements here, other than contributing the motivation to features on the unnamed functional projections F. Further challenges to this type of analysis are raised by corpus data like (35):

  1. (35)

    1. (a) He would not cut [half [[so good] a figure]].

    2. (b) I will not be the medium of [any [[so absurd] a requisition]].

    3. (c) Is there [any [such a thing]] as a precision strike at North Korea's nuclear capacity?

Within a functional projection-based approach, the presence of an additional element before the BMC requires one additional position before the FP. As a solution, following Wood (Reference Wood2002) and Wood & Vikner (Reference Wood, Vikner, Sleeman and Perridon2011), one could adopt another modified functional projection hierarchy as in (36):Footnote 11

  1. (36)

Within this revised functional projection, the derivation now moves the DegP to the Spec of NumP. Since there is still one more layer over this NumP, we could predict the appearance of determiners like any. By having one empty slot in the Spec of DP, we might expect one more additional expression preceding any or no.

As noted in passing by Kennedy & Merchant (Reference Kennedy and Merchant2000), this kind of movement (SpecFP or SpecNumP) is driven by some features like [+wh]; but we have seen that not all attributive modifiers carry such [+wh] features. Even if we can have a way of dealing with this issue, questions still remain of how to predict what licenses the phrases which can front, as noted earlier.

We have also seen that each degree adverb displays different lexical idiosyncrasies. For example, comparatives such as more/less allow optional inversion:

  1. (37)

    1. (a) That's [a steeper an angle] than it's really going to be. (COCA SPOK CBS_SunMorn)

    2. (b) He has become [a more disciplined hitter] and a smarter hitter. (COCA NEWS NewYorkTimes)

Whatever feature triggers the DegP inversion, the one that more/less carries is different from the others (cf. Troseth Reference Troseth and den Dikken2004). In addition, observe the behaviour of enough:

  1. (38)

    1. (a) If that's not [a good enough excuse], he has others. (COCA NEWS USAToday)

    2. (b) And betrayed is not [strong enough a word]. (COCA SPOK CNN_ KingWknd)

The analysis has to guarantee both the movement of the degree item enough in the postadjectival position and then movement of the word in the Spec of a higher functional projection. It is hard to find any syntactic motivations for such bidirectional movements, other than positing a purely constructional reason (cf. Culicover & Jackendoff Reference Culicover and Jackendoff2005; Goldberg Reference Goldberg2006).

In dealing with the similarities and differences between so and such, Wood & Vikner (Reference Wood, Vikner, Sleeman and Perridon2011) provide an interesting movement approach.Footnote 12 In particular, they propose that both the so + AP and such are generated in the predicate position:

  1. (39)

This predicate-driven movement is claimed to reflect the property that the attributive AP in the BMC basically functions as a predicative phrase (data from Wood & Vikner Reference Wood, Vikner, Sleeman and Perridon2011):

  1. (40)

    1. (a) the shoplifter is ashamed

    2. (b) *the ashamed shoplifter

    3. (c) *the so ashamed shoplifter

    4. (d) so ashamed a shoplifter

Attractive though this analysis appears to be, there are some issues it needs to face. Adjectives like inner, mere and wooden have only attributive uses:

  1. (41)

    1. (a) There was a low, wooden form fixed to the wall. vs *The wall is wooden.

    2. (b) Belief is such an intensely inner matter. vs *The matter is inner.

    3. (c) I cannot pay a mere child more than one dollar. vs *The child is mere.

The predicate analysis in which the inverted AP is generated in the predicate position would not then license such non-predicative attributive-only adjectives in the BMC. However, our corpus search reveals counterexamples to this prediction:

  1. (42)

    1. (a) Without so much as a waving-line in them, it becomes [so wooden a form].

    2. (b) It may seem a strange requirement in [so inner a matter] as friendship that one should insist upon the objective mood as fundamental.

    3. (c) I boxed a little boy, a nephew of mine, very unrighteously, and he is so little, [so mere a child], that I can't ask his pardon.

The movement-based approaches we have discussed here provide some insights into the complexity of the BMC, in particular in relation to the ordering of the elements involved. However, the properties of degree expressions, APD and BMC indicate that an array of functional projections and movement processes are not enough to capture their flexibility, lexical idiosyncrasies and constructional constraints.

5 A lexicalist, construction-based approach

5.1 HPSG: a brief background

HPSG uses typed feature structures to model linguistic objects. Feature structures of various types specify values for appropriate features. Some of these values are themselves complex feature structures, as, for example, in a word or phrase:Footnote 13

  1. (43)

The elements in the arg-st (argument-structure) will be realized as subj (subject) and comps (complements) in syntax in accordance with the Argument Realization Constraint:Footnote 14

  1. (44)

This realization constraint has the arg-st at the lexical level linked to the valence features subj and comps, which are syntax-sensitive.

In the version of HPSG that we assume here, complex phrases are licensed by grammatical constructions, which are schemata imposing constraints on how component signs can combine to build larger signs. Two constructions of English will suffice for our purposes here, the head–subject construction and the head–complement construction, given both in the form of Pollard & Sag's (Reference Pollard and Sag1994) schemata and the construction types of Fillmore (Reference Fillmore, Webelhuth, Koenig and Kathol1999), Sag et al. (Reference Sag, Wasow and Bender2003), Kim & Sells (Reference Kim and Sag2008), Sag (Reference Sag, Boas and Sag2010) and related work:

  1. (45)

    1. (a) Head–Subject Construction:

    2. (b) Head–Complement Construction:

The Head–Subject Construction in (45a) allows the combination of a VP with its subject, whereas the Head–Complement Construction in (45b) licenses the combination of a lexical head and its phrasal complements. These constructions interact with the Head Feature Principle and the various (partly parochial) linear precedence (LP) constraints to license complex phrasal signs:Footnote 15

  1. (46) The Head Feature Principle:

    In a headed construction, the head value of the mother must be identical to the head value of the head daughter.

The well-formed signs defined by our grammar are those that instantiate the mother of some construction, obeying all constraints specified by that construction, all general principles and all linear precedence constraints. We can represent the construction of complex expressions via trees of a familiar sort, which are simply a convenient representation of how a given complex sign (a phrase) is licensed by our grammar. (47) is one such diagram:

  1. (47)

The verb petted selects two arguments as given in the arg-st value. The first one is realized as subj and the second one as comps value. The Head Feature Principle guarantees that the head feature of the verb is eventually identified with that of the sentence. The ordering among constituents is constrained by the LP constraints specifying that the lexical head precedes its complement, whereas the VP follows the subject.

5.2 Head–functor construction

Our analysis, following Van Eynde (Reference Van Eynde and Müller2007), starts with the observation that there are many similarities between specifiers and modifiers in languages. For example, in Italian, the same agreement marker appears on specifiers as well as modifiers (data from Van Eynde Reference Van Eynde and Müller2007):

  1. (48)

Treating specifiers and modifiers in a uniform way as Van Eynde (Reference Van Eynde and Müller2007) and Kay & Sag (Reference Kay and Sag2009) do, we assume that specifiers and modifiers are both functors that ‘select’ their head. More specifically, we accept the view that English employs the head–functor construction in (49) as one of the well-formed phrasal combinations, as represented in the following:

  1. (49) Head–Functor Construction (first version):

What this constructional constraint specifies is that the combination of a head and a functor selecting this head forms a well-formed hd-functor-cx. In English, various functor elements can serve as non-head daughters in a local tree and ‘select’ their head sister through the feature sel (select). The feature specifies what kind of head a functor (encompassing both modifier and specifier) can combine with in syntax.Footnote 16 Examples like the following are all head–functor combinations in which either a modifier or a specifier combines with its semantic head argument:

  1. (50)

    1. (a) [[F big] [H mess]]

    2. (b) [[F the] [H big mess]]

    3. (c) [[F all] [H the students]]

As an illustration, consider the structure of (50c):

  1. (51)

As shown here, both all and the are functors that select an argument head. The functor the first combines with the head students, and the resulting phrase then serves as the sel value of the functor all. Both combinations are well-formed head–functor constructions.

With the postulation of the Head–Functor Construction as one of the well-formed English phrasal types, we can account for the complex distributional possibilities of such. In dealing with the distribution of such, we first need to consider the fact that such cannot combine with a definite NP, as in *such the big mess, though it is allowed to combine with an indefinite NP, as in such occasions. This means that functors like such require the sister to carry a special marker value originating from an expression like a (in contrast to the). For this purpose, following Van Eynde (Reference Van Eynde and Müller2007), we introduce the feature mrk (marker) whose value is passed up to the mother from a non-head daughter.Footnote 17

We also need to consider that, as noted earlier, such has flexible distributional possibilities. For example, such can also appear in the determiner or adjectival position (cf. Wood Reference Wood2003; Kim & Kim Reference Kim and Kim2009), in addition to the predeterminer position:

  1. (52)

    1. (a) I would imagine that nobody has done such a thing. <ICE-GB:S1B-023 005:1>

    2. (b) I'm normally oppressed by such paintings. <ICE-GB:S1B-018 171:1>

    3. (c) Many such parents will be tempted to wash their hands of their responsibilities. <ICE-GB:W2C-007 080:2>

All these three positions can be taken as ‘adnominal’ in the sense that they appear in the prenominal position. As a way of expressing this observation, rather than assigning three different lexical categories to each of these three different cases, we introduce the category adnominal as a supertype, shown in the following hierarchy of parts of speech:

  1. (53)

This multiple inheritance hierarchy, adopted from Kim & Sag (Reference Kim and Sells2005), is meant to capture many similarities between lexical categories (cf. Flickinger & Nerbonne Reference Flickinger and Nerbonne1992). As is well known, NP and CP behave like nominal elements, whereas CP and VP are verbal in the sense that they both denote a propositional meaning. Predeterminers, adjectives and determiners behave alike in many languages and are treated as belonging to the same category adnominal.Footnote 18

To reflect the combinatorial constraints and distributional possibilities we have just observed, we assume that the word such belongs to the supercategory adnominal with an Nʹ as its sel value:Footnote 19

  1. (54)

The lexical information indicates that such can select either an unmarked Nʹ or an indefinite NP. The feature mrk picks out some property of a word or a phrase which other aspects of the constructions may be sensitive to. In most cases, the value of mrk is unmrk (unmarked).Footnote 20 The simple lexical entry in (54) would then first project structures like the following:

  1. (55)

    1. (a)

    2. (b)

In the structure (55a), both such and a are functors selecting an indefinite NP and an unmarked Nʹ, respectively. (55b) is similar in that such forms a head–functor–cx with its head phrase good parents. However, due to the tight restriction on the mrk value, the analysis would not license phrases such as these:

  1. (56) *such the big mess, *such this dog, *such my dog, *such many dogs, *such any dog, . . .

These are all unacceptable simply because the functor such would have to combine with an NP whose mrk value is def (definite), as shown in (57), in violation of the lexical entry for such (see (54)):

  1. (57)

The present analysis can also be extended to other peculiar distributions of such:Footnote 21

  1. (58)

    1. (a) There was no foreseeable possibility of any such a scheme.

      <BNC A8X 960>

    2. (b) Without promotion, there is no such a thing anymore.

      <BNC HCX 352>

Such examples indicate that the combination of the functor such with its head does not close off the NP projection. To generate such combinations, the only thing we need to assume is that such has a mrk value unmrk and that determiners like no or any select for nominal expressions which are unmrk:

  1. (59)

This approach will rule out examples like *no a man and *no the man: both a and the have their own mrk value passed up to the mother and the resulting phrase a man or the man will then have mrk values a and the respectively.

The lexical information of such, interacting with the constructional properties of the Head–Functor Construction, can therefore encompass the flexible distributions of such (cf. Siegel Reference Siegel1994; Seppänen et al. Reference Seppänen, Granath and Danielsson2002; Spinillo Reference Spinillo2003).Footnote 22 However, the closely related what is actually quite different from such, and it only appears in the predeterminer position. For example, what can combine with an indefinite NP, bare countable or uncountable Nʹ:

  1. (60)

    1. (a) The fans can understand what [a great guy] he is. (COCA NEWS Atlanta)

    2. (b) What [remarkable children] you are. (COCA NEWS SanFrancisco)

    3. (c) What [wonderful food] it is. (COCA MAG MotherEarth)

We can assign the correct lexical properties to what in order for it to be a functor, licensing such examples:Footnote 23

  1. (61)

As shown here, the NP a great guy has a non-empty mrk value. The functor what combines with this head NP, forming a head–functor construction. Note that the resulting phrase also has a mrk value inherited from the functor what. This also means that a structure like (61) could not be an argument of a functor such as no or any, which requires a nominal expression with no mrk value:

  1. (62) *any what a great guy, *no what remarkable children, no a great guy,

As seen in (59), determiners like any or no require the head NP to have no mrk value. The interaction with these determiners shows us that the resulting combination of the functor what and its head is different from the combination of such and its head. At this finer level of detail, this entails that the such-type BMC needs to be differentiated into two subtypes with different constraints on the resulting phrase, displaying their lexical idiosyncrasies.

5.3 On the so-type BMC

The behaviour of the so-type BMC is partly similar to that of the such-type BMC, but there are several differences. One main difference is that so first combines with an AP and then an indefinite NP as represented in the following structure:Footnote 24

  1. (63)

In this structure, the APD so big and the NP a mess are both well-formed head–functor constructions. The question then arises as to what licenses the combination between the AP so big and the indefinite NP a mess. We could, as Van Eynde (Reference Van Eynde and Müller2007) and Kay & Sag (Reference Kay and Sag2009) do, posit a special construction that licenses this combination.Footnote 25 However, we take a slightly different direction. Rather than assigning this combinatorial power to a construction at the phrasal level, we attribute the combinatorial possibilities to the lexical properties of degree words so or such. Recall that both so- and such-type degree words can appear in a (immediate or non-immediate) predeterminer position:

  1. (64)

    1. (a) so-type: so, as, too, this, that, more, how(ever)

      as good a singer, how strange a story, too hot a day, how(ever) brave a soldier, far cheaper a method, . . .

    2. (b) such-type: what, many, half, . . .

      many a time, such a disgrace, what a pity, half an hour, . . .

There is no doubt that the degree words play a crucial role in licensing these combinations. In addition, consider the following simple variations of so and such with a DCM again:

  1. (65)

    1. (a) The industry is so big a mess (that no one can control it).

    2. (b) The industry is such a big mess (that no one can control it).

If we consider the so- and such-types of examples, the question arises of whether a uniform analysis for both is possible. As we have seen, so also has much more flexible, distributional possibilities even with a DCM:

  1. (66)

    1. (a) The music is so loud (that I can't sleep).

    2. (b) The gentleman spoke so quickly (that I couldn't understand him).

Here so does not participate in a structure with an NP dependent after the AP or AdvP, but directly combines with an optional DCM, showing us that this must be one of its lexical properties.

In order to capture the necessary generalities as well as being able to determine idiosyncratic properties, our analysis minimizes the postulation of phrasal constructions but places the combinatorial power in the lexical properties of so and such. In addition, we assume that the functor can select more than one element – including a DCM – as represented in the following lexical information:

  1. (67)

Slightly different from such, the adverbial functor so can select one obligatory dependent together with two optional dependents including a DCM. This reflects the fact that while the DCM is optional and discontinuous, the presence of the DCM is dependent upon the degree word which licenses it.Footnote 26 In the present analysis, the combination of the AP and the following NP is thus not a constructional constraint, but is licensed by the lexical properties of the degree item so. This lexical specification, together with an assumption that the functor combines with one head element at a time, would license the following binary-branching structure:

  1. (68)

The functor so, selecting three dependents, first combines with its AP head, discharging the first AP requirement of the sel feature. The result will then combine with the NP head a mess, forming a bigger Head–Functor Construction. Then, this once again combines with the third dependent DCM clause, fulfilling the sel requirement.

It may be observed that the combinations here are controlled not by the head but by the functor, and the value of the sel value is passed up to the mother from the non-head daughter until it is discharged. This requires us to slightly revise the Head–Functor Construction in (49) as follows:

  1. (69) Head–Functor Construction (second version):

The notation here designates a list, while is a subtraction operation on the list. This revised construction dictates that the value of the sel feature is passed up to the mother, minus the discharged part. This is stated in words in (70):

  1. (70) Head–Functor Construction:

    A phrase's sel value is identical with that of a functor (non-head) daughter minus the discharged value.

The revised constraint guarantees that the unsaturated sel value is passed up to the mother.

As a reviewer points out, examples like (very) few houses require further elaboration of our constraint. In such examples, it is few that requires a plural head noun houses. This means that when the head few combines with the functor very, the sel feature that it carries needs to pass up to the mother. This is one main reason why Van Eynde (Reference Van Eynde and Müller2007) treats the sel feature as a head feature. In the present analysis, to allow the sel value of the functor as well as that of the head daughter to pass up to the mother phrase, we need to slightly revise (70), as in the following:Footnote 27

  1. (71) Head–Functor Construction (final version):

    A phrase's sel value is identical with the combination of its functor's and head daughter's sel value minus the discharged value.

The features mrk and sel are both inherited from functors, but the latter is different in the sense that the value from the mother is also passed to the mother. This mechanism, reminiscent of foot features in GPSG and argument composition in HPSG, thus allows the certain values in head or non-head daughters to pass up to the mother (see Gazdar Reference Gazdar1981 for foot features and Abeillé et al. Reference Abeillé, Godard, Sag, Hinrichs, Kathol and Nakazawa1998 for argument composition).

The present analysis allows degree expressions like so and such to carry more lexical restrictions on their possible dependents. One difference between the two, mentioned above, is that the former can place the restriction on the first NP dependent after forming a phrasal combination with an AP, while such imposes its constraints directly:

  1. (72)

    1. (a) *[[so big] [the mess]], *[[so big] [many the mess]], *[[so big] [somewhat mess]], . . .

    2. (b) *[[such] [the big mess]], *[[such] [many the big mess]], *[[such] [somewhat big mess]], . . .

In addition, note that unlike such, so does not combine with a bare NP; nor can it combine with plural NPs:

  1. (73)

    1. (a) You have been [*so/such good friends] to the students.

    2. (b) You are in [*so/such good shape].

These reflect lexical differences between the two degree words.

As noted above, the so-type APD can also be followed by the preposition of. Our corpus search indicates that the most frequent cases of of-insertion are with the APD trigger more. Other triggers such as so, too, as, too, this and that are also possible, but less frequent:

  1. (74)

    1. (a) It is [more simple of a tip]. (COCA NBC today)

    2. (b) Maybe I am not [as good of a parent] as I thought I was. (COCA SPOK CBS_sixty)

In the present system, the only thing we need to do is to modify the lexical entry slightly:

  1. (75)

As shown in the lexical entry, the APD trigger can select either an indefinite NP or a PP with the meaningless of. The preposition of is optional, but a is obligatory. This lexical information will project a structure like the following:Footnote 28

  1. (76)

As noted by Wood & Vikner (Reference Wood, Vikner, Sleeman and Perridon2011), predicative adjectives are preferred in the BMC. This may suggest that the preposition of is introduced to mark the fact that the prepositional object is the complement of the adjectival element. This structure can then be interpreted as reflecting the property such that when a degree item like more selects an AP, it can further require a dependent NP or a PP with of with no meaning difference. Note that even with the PP dependent, its object NP still needs to observe the indefiniteness constraint:

  1. (77)

    1. (a) It is more simple of a/*the trip.

    2. (b) Maybe I am not as good of a/*the parent as I thought I was.

As we have noted earlier, there is another variation which needs to be lexically specified, supporting the direction of our analysis. For example, the degree items more and enough can either be in the APD or used in the postdeterminer position:

  1. (78)

    1. (a) more a pragmatic approach vs a more pragmatic approach

    2. (b) big enough a house vs a big enough house

This is variation in the sel value in the present analysis. That is, more or enough will lexically select either an Nʹ (pragmatic approach) or a full NP (a pragmatic approach). In addition, note that more does carry a mrk value, passing it up to the mother phrase. In section 5.2, we saw that expressions like no or any can combine with any unmarked nominal expression. The combined possibilities from these lexical properties are illustrated by corpus examples like the following:

  1. (79)

    1. (a) There is no more noble a sacrifice. (COCA NEWS NewYorkTimes)

    2. (b) There is no more serious a task for Congress than deciding whether to impeach a sitting president. (COCA NEWS Houston)

In these examples, more will form a BMC first, after combining with noble a sacrifice, and this BMC phrase is what the functor no selects. The lexical properties, together with the combinatorial possibilities of the head–functor construction, can explain variation in the formation of BMCs.

5.4 More on APD and DCM together

As we noted earlier, some of the so-type and such-type BMCs can also be linked to another syntactic expression, such as an as- or that-clause in a discontinuous position, which we call a DCM:

  1. (80)

    1. (a) It is also [so] [dark] [a cave] [that we could not see a thing].

    2. (b) It is [so] [intense] [a light] [as to blind the eyes].

  2. (81)

    1. (a) It was [such] [a good job] [that I would never have quit].

    2. (b) He does [such] [a good job] [that it hardly seems worthwhile to compete].

    3. (c) You don't have half [such] [a hard time] [as I do].

These examples indicate that the functor so and such can select more than one element which functions as a modifier. The type of this DCM is also restricted:

  1. (82)

    1. (a) *It is also [so] [dark] [a cave] [which we could not see a thing].

    2. (b) *It was [such] [a good job] [which I would never have quit].

As noted earlier, this dependency intuitively originates from the degree words so and such, not from a construction such as the APD. The present analysis can provide a simple way of stating this dependency, allowing the degree items to select one more dependent:

  1. (83)

This kind of lexical information will project a structure like the following for (81a):

  1. (84)

As shown here, such has two dependents: an indefinite NP and a CP. It first combines with the indefinite NP a mess, forming a head–functor phrase. The result will once again combine with the resulting CP clause. We treat the relationship between the BMC and the DCM not as a type of movement or dislocated phenomenon but as a locally bound construction.

6 Conclusion

We have attributed the discontinuous properties of so- and such- type degree expressions to their lexical properties, rather than to independent constructions. Their behaviour is the result of interactions between their lexical properties and the properties of more general constructions. This way of looking at the two types of BMC has allowed us to capture the properties of so- and such-type expressions in a uniform, lexicalist way. In particular, the BMCs are special in the sense that they select two (or more) arguments. In addition to head–complement relations, head–functor relations also play an important role in grammar. In our analysis, some discontinuous constructions are not really discontinuous, but are selected by the degree elements.

The analysis we sketched reflects the following basic observations:

  • • Intuitively, the degree adverbs determine what to combine with; they are functors.

  • • So- and such-type degree expressions are not different as functors; but they are different in what they select as their arguments.

  • • A functor can select more than one argument, as a head does. The value of the feature sel is a list (a sequence of values).

  • • A head–functor construction is a binary structure in which a functor combines with its arguments one by one.

Interactions among constructional and lexical constraints provide us with a way of capturing complex phenomena including the DCM. Every language employs a certain number of constructions which cannot be predicted only from general grammar rules. Based on our in-depth corpus search, it is clear that in a full-scale analysis we need to pay close attention to the trade-off between very specific lexical properties and constructions with more generality.

Footnotes

2 The term originates from Berman (Reference Berman1974).

3 The corpora we use in this study include the ICE-GB (International Corpus of English, Great Britain), COCA (Corpus of Contemporary American English) and BNC (British National Corpus).

4 As noted by Wood & Vikner (Reference Wood, Vikner, Sleeman and Perridon2011), such and so are historically related: such in Old English was swylc from a Proto-Germanic compound *swalikaz ‘so formed’ (swa ‘so’ + *likan ‘form’).

5 We use the term APD to refer to the predeterminer expression while using the ‘BMC’ to encompass the APD plus the following NP.

6 Canonical discontinuous modifiers (DCM) include examples like (i) in which the dependent elements are separated by a head noun:

  1. (i)

    1. (a) That's a [difficult] question [to answer]. (BNC JK9 S_unclassified)

    2. (b) The same may also be true of a [similar] approach [to curriculum development]. (BNC FAM W_ac_polit_law_edu)

The infinitival VP in (ia) is linked to the adjective difficult, whereas the PP in (ib) is to the adjective similar. In both cases, the two dependent elements are intervened by the head noun. See Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan1999) and de Mönnink (Reference de Mönnink, Percy, Meyer and Lancashire1996, Reference de Mönnink and Kirk2000) for further types of the DCM and their distributional properties.

7 The notation ø means that no article is realized in this position.

8 When so is used as an (adverbial) conjunct, it can be followed by a plural NP:

  1. (i)

    1. (a) That's exactly why I painted them. So other people can appreciate them. (COCA FIC Triquarterly)

    2. (b) I see it more as a need for job training, so young men can connect their work with a reward. (COCA MAG SportsIll)

9 Bolinger (Reference Bolinger1972) assumes such has basically two semantic functions: intensifying and identifying. See also Wood (Reference Wood2002) and Spinillo (Reference Spinillo2003).

10 The DCM element is optional, as it functions as a modifier.

11 See Ritter (Reference Ritter1992), Matushansky (Reference Matushansky2002) and Alexiadou et al. (Reference Alexiadou, Haegeman and Stavrou2007) for some motivations for introducing the NumP.

12 These authors discuss the internal and external syntax of these two etymologically related words so and such in English, Danish and German.

13 The feature structures used here are abbreviated. See Sag et al. (Reference Sag, Wasow and Bender2003) and Kim & Sells (Reference Kim and Sag2008).

14 The constraint specifies that the first element in the arg-st is realized as subj and the remaining values as the comps. See Sag et al. (Reference Sag, Wasow and Bender2003) and Kim & Sells (Reference Kim and Sag2008).

15 See Sag et al. (Reference Sag, Wasow and Bender2003) for detailed discussion of the linear precedence constraints.

16 The feature sel is different from canonical valence features such as sub and comps in HPSG. In addition, differently from Van Eynde (Reference Van Eynde and Müller2007) and Sag (2009), the feature sel as we articulate it here is not a head feature, but a non-head feature passed up to the mother when it is not discharged.

17 Motivations for introducing the feature mrk and related constraints can be found in Van Eynde (Reference Van Eynde and Müller2007) and Kay & Sag (Reference Kay and Sag2009), as well as in our analysis below.

18 In languages like Korean, for example, adjectives and determiners can both act like a modifier in the prenominal position, showing no ordering restrictions and no complementary distribution:

  1. (i)

    1. (a) chakhan ku haksayng

      honest the student

      ‘the honest student’

    2. (b) ku chakhan haksayng

      the honest student

      ‘the honest student’

We assume that English also needs a lexical supertype adnominal encompassing both adjectives and determiners.

19 The notation | in the sel value means that such can select either an Nʹ or a full NP whose mrk value is a. We could simplify this sel value by introducing a meta-marker value like nondefinite which has two subtypes unmrk and a. This would then enable us to say such selects a nominal element whose mrk value is only nondefinite.

20 The mrk value, which is a non-head feature, plays an important role in specifying the value of determiner. See Van Eynde (Reference Van Eynde and Müller2007) for further constraints on mrk.

21 See Altenberg (Reference Altenberg, Oostdijk and de Hann1994) for detailed discussion of the functions of such.

22 As a reviewer points out, only a limited range of determiners can be in this Det position; *this such a mess is not possible, for example. This in turn means that the determiner no lexically requires an unmarked Nʹ as its head expression.

23 The analysis sketched here is similar to Van Eynde (Reference Van Eynde and Müller2007).

24 As noted in Van Eynde (Reference Van Eynde and Müller2007), another possible structure is the one where the indefinite article selects the APD as its specifier, and the resulting phrase (DetP) serves as the specifier of a common noun. One problem that such an analysis encounters lies in classical constituency tests in Never before had we seen [that big] [a bridge] vs *Never before had we seen [bridge] [that big a]. This contrast indicates that the APD and the following indefinite article do not form a constituent. See Van Eynde (Reference Van Eynde and Müller2007) for further discussion of the possible problems facing such an analysis.

25 Van Eynde (Reference Van Eynde and Müller2007) posits a special construction to license an AP with the functor so to combine with an indefinite NP. Kay & Sag (Reference Kay and Sag2009), slightly differently from Van Eynde (Reference Van Eynde and Müller2007), introduce a construction apd-cxt (adjectival predeterminer construction). Simplifying slightly, the phrase so big has a constructional property which adds an indefinite NP as its sel value.

26 Kay & Sag (Reference Kay and Sag2009) treat the DCM dependency as an extraposition process. Since the DCM comes immediately after the BMC, we treat it as a lexical dependency. See also de Mönnink (Reference de Mönnink and Kirk2000) and Müller (Reference Müller2004).

27 A formalization of this construction could be as in (i):

  1. (i) Head–Functor Construction:

As this feature structure indicates, the mother's sel value is the union (marked by the list operation ⊕) of the non-head functor daughter's sel value and the head daughter's sel value, minus the discharged value. This discharged value () is the head daughter itself.

28 Bennis et al. (Reference Bennis, Corver and den Dikken1998) take of and a to occupy the same head, comparing the Dutch and English N of N construction.

References

Abeillé, Anne, Godard, Daniele & Sag, Ivan A.. 1998. Two kinds of composition in French complex predicates. In Hinrichs, Erhard, Kathol, Andreas & Nakazawa, Tsuneko (eds.), Complex predicates in nonderivational syntax, 1–41. San Diego: Academic Press.Google Scholar
Abney, Steven. 1987. The English noun phrase in its sentential aspect. PhD dissertation, MIT.Google Scholar
Alexiadou, Artemis, Haegeman, Lilianne & Stavrou, Melita. 2007. Noun phrase in the generative perspective. Berlin: Mouton de Gruyter.CrossRefGoogle Scholar
Altenberg, Bengt. 1994. On the function of such in spoken and written English. In Oostdijk, Nelleke & de Hann, Pieter (eds.), Corpus-based research into language, 223–39. Amsterdam: Rodopi.CrossRefGoogle Scholar
Bennis, Hans, Corver, Norbert & den Dikken, Marcel. 1998. Predication in nominal phrases. Journal of Comparative Germanic Linguistics 1, 85–117.CrossRefGoogle Scholar
Berman, Arlene. 1974. Adjectives and adjective complement constructions in English. PhD dissertation, Harvard University.Google Scholar
Biber, Douglas, Johansson, Stig, Leech, Geoffery, Conrad, Susan & Finegan, Edward. 1999. Longman grammar of spoken and written English. London: Longman.Google Scholar
Bolinger, Dwight. 1972. Degree words. The Hague: Mouton de Gruyter.CrossRefGoogle Scholar
Bresnan, Joan W. 1973. Syntax of the comparative clause construction in English. Linguistic Inquiry 4 (3), 275–343.Google Scholar
Corver, Norbert. 1990. The syntax of left branch constructions. PhD dissertation, Tilburg University.Google Scholar
Corver, Norbert. 1997. Much-support as a last resort. Linguistic Inquiry 28, 119–64.Google Scholar
Corver, Norbert. 2002. On three types of movement within the Dutch nominal domain. In Coene, Martine & D'hulst, Yves (eds.), From NP to DP, vol. 1: The syntax and semantics of noun phrases, 297–328. Amsterdam: John Benjamins.Google Scholar
Culicover, Peter & Jackendoff, Ray. 2005. Simpler syntax. Oxford: Oxford University Press.CrossRefGoogle Scholar
de Mönnink, Inge. 1996. A first approach to the mobility of noun phrase constituents. In Percy, Carol E., Meyer, Charles F. & Lancashire, Ian (eds.), Synchronic corpus linguistics: Papers from the 16th International Conference on English Language Research on Computerized Corpora, 143–57. Amsterdam: Rodopi.Google Scholar
de Mönnink, Inge. 2000. A moving phrase: A multi-method approach to the mobility of constituents in the English noun phrase. In Kirk, John M. (ed.), Corpora galore: Analyses and techniques in describing English, 133–47. Amsterdam: Rodopi.CrossRefGoogle Scholar
Fillmore, Charles. 1999. Inversion and constructional inheritance. In Webelhuth, Gert, Koenig, Jean-Pierre & Kathol, Andreas (eds.), Lexical and constructional aspects of linguistic explanation: Studies in constraint-based lexicalism, 113–28. Stanford: CSLI Publications.Google Scholar
Flickinger, Daniel & Nerbonne, John. 1992. Inheritance and complementation: A case study of easy adjectives and related nouns. Computational Linguistics 19 (3), 269–309.Google Scholar
Gazdar, Gerald. 1981. Unbounded dependencies and coordinate structure. Linguistic Inquiry 12, 155–84.Google Scholar
Giorgi, Alessandra & Longobardi, Giuseppe. 1991. The syntax of noun phrases. Cambridge: Cambridge University Press.Google Scholar
Goldberg, Adele. 2006. Constructions at work. Oxford: Oxford University Press.Google Scholar
González Escribano, José Luis. 2005. Discontinuous APs in English. Linguistics 43 (3), 563–610.Google Scholar
Haegeman, Liliane & Guéron, Jacqueline. 1999. English grammar: A generative perspective. Oxford: Blackwell.Google Scholar
Hendrick, Randall. 1990. Operator movement in NP. In Halpern, Aaron L. (ed.), Proceedings of the 9th West Coast Conference on Formal Linguistics, 249–64. Stanford: CSLI Publications.Google Scholar
Huddleston, Rodney & Pullum, Geoffrey et al. 2002. The Cambridge grammar of the English language. Cambridge: Cambridge University Press.CrossRefGoogle Scholar
Kay, Paul & Sag, Ivan A.. 2009. Not as hard a problem to solve as you might have thought. Proceedings of the HPSG-09 Conference. Stanford: CSLI Online Publications.Google Scholar
Kennedy, Christopher & Merchant, Jason. 2000. Attributive comparative deletion. Natural Language and Linguistic Theory 18, 89–146.CrossRefGoogle Scholar
Kim, Jong-Bok & Kim, Kyeongmin. 2009. A functor treatment of such: A corpus-based approach. Korean Journal of Linguistics 33 (4), 853–70.Google Scholar
Kim, Jong-Bok & Sag, Ivan A.. 2005. Variations in English object extraposition. Proceedings of the 41st Annual Meeting of the Chicago Linguistic Society, 192–212. Chicago: CLS.Google Scholar
Kim, Jong-Bok & Sells, Peter. 2008. English syntax: An introduction. Stanford: CSLI Publications.Google Scholar
Lechner, Winfried. 1999. Comparatives and DP-structure. PhD dissertation, University of Massachusetts.Google Scholar
Matushansky, Ora. 2002. Movement of degree/degree of movement. PhD dissertation, MIT.Google Scholar
Merchant, Jason. 1999. The syntax of silence: Sluicing, islands, and identity in ellipsis. PhD dissertation, University of California.Google Scholar
Moro, Andrea. 1997. The raising of predicates: Predicative noun phrases and the theory of clause structure. Cambridge: Cambridge University Press.CrossRefGoogle Scholar
Müller, Stefan. 2004. Continuous or discontinuous constituents? A comparison between syntactic analyses for constituent order and their processing systems. Research on Language and Computation 2 (2), 209–57.CrossRefGoogle Scholar
Pollard, Carl & Sag, Ivan A.. 1994. Head-driven phrase structure grammar. Stanford: CSLI Publications.Google Scholar
Quirk, Randoph, Greenbaum, Sidney, Leech, Geoffrey & Svartvik, Jan. 1985. A comprehensive grammar of the English language. London: Longman.Google Scholar
Ritter, Elizabeth. 1992. Cross-linguistic evidence for number phrase. Canadian Journal of Linguistics 37, 197–218.CrossRefGoogle Scholar
Sadler, Louisa & Arnold, Douglas J.. 1994. Prenominal adjectives and the phrasal/lexical distinction. Journal of Linguistics 30, 187–226.CrossRefGoogle Scholar
Sag, Ivan A. 2010. Sign-based Construction Grammar: An informal synopsis. In Boas, Hans & Sag, Ivan A. (eds.), Sign-based Construction Grammar. Stanford: CSLI Publications.Google Scholar
Sag, Ivan A., Wasow, Tom & Bender, Emily. 2003. Syntactic theory: An introduction. Stanford: CSLI Publications.Google Scholar
Seppänen, Aimo, Granath, Solveig & Danielsson, Lars. 2002. The construction ‘AdjP–a(n)–Noun’ in Present-day English syntax. Leuvense Bijdragen 91, 97–136.Google Scholar
Siegel, Muffy. 1994. Such: Binding and the pro-adjective. Linguistics and Philosophy 17, 481–97.CrossRefGoogle Scholar
Spinillo, Mariangela. 2003. On such. English Language and Linguistics 7 (2), 195–210.CrossRefGoogle Scholar
Troseth, Erika. 2004. Negative inversion and degree inversion in the English DP. In den Dikken, Marcel (ed.), Linguistics in the Big Apple: CUNY/NYU working papers in linguistics, 1–15.Google Scholar
Van Eynde, Frank. 2006. NP-internal agreement and the structure of the noun phrase. Journal of Linguistics 42 (1), 139–86.CrossRefGoogle Scholar
Van Eynde, Frank. 2007. The Big Mess Construction. In Müller, Stefan (ed.), The proceedings of the 14th International Conference on Head-driven Phrase Structure Grammar, 415–33. Stanford: CSLI Publications.Google Scholar
Vikner, Sten. 2001. Predicative adjective agreement. In Adamzik, Kirsten & Christen, Helen (eds.), Sprachkontakt, Sprachvergleich, Sprachvariation: Festschrift für Gottfried Kolde, 399–414. Tübingen: Niemeyer.CrossRefGoogle Scholar
Wood, Johanna. 2002. Much about such. Studia Linguistica 56 (1), 91–115.CrossRefGoogle Scholar
Wood, Johanna. 2003. Number phrase and fronted pre-modifiers in Middle English. Proceedings of the 32nd Western Conference on Linguistics, 305–18. Fresno, CA: Dept of Linguistics, CSU.Google Scholar
Wood, Johanna & Vikner, Sten. 2011. Noun phrase structure and movement: A cross-linguistic comparison of such/sådan/solch and so/så/so. In Sleeman, Petra & Perridon, Harry (eds.), The noun phrase in Romance and Germanic: Structure, variation, and change, 89–110. Amsterdam: John Benjamins.CrossRefGoogle Scholar
Zamparelli, Roberto. 1995. Layers in the determiner phrase. PhD dissertation, University of Rochester.Google Scholar
Zwicky, Arnold. 1994. Exceptional degree markers: A puzzle in internal and external syntax. OSU Working Papers in Linguistics 47, 111–23.Google Scholar