A Cleft Palette - Graduate Center, CUNY · A Cleft Palette On the lands cape of cleft co nstr uctio...

A Cleft PaletteOn the landscape of cleft constructions and their syntactic derivationsCleft Workshop 2008 • Zentrum für Allgemeine Sprachwissenschaft (ZAS) • Berlin • 29 November 2008Marcel den Dikken — Linguistics Program — CUNY Graduate Center — [email protected]

1 A typology of cleft sentences

1.1 Pseudoclefts: Predicational vs specificational

(1) a. what Johni is is important to him i

6 PREDICATIONAL: important to him predicates a property of what John is

b. what Johni is is important to himselfi

6 SPECIFICATIONAL: important to himself specifies a ‘VALUE’ for the variable in the wh-clause

6 for (1a), the term ‘predicational pseudocleft’ is really a misnomer: there is nothing ‘(pseudo)cleft’

about this kind of sentence — it represents a garden-variety predicational copular sentence, with the

free relative as the subject and the postcopular AP as the predicate

6 for (1b), the term ‘specificational pseudocleft’ is entirely apt — (1b) ‘cleaves’ the sentence John is

important to himself, with important to himself specifying a value for the variable in the wh-clause

6 in specificational pseudoclefts, the postcopular constituent seems to be the underlying subject of the

predicate denoted by the wh-clause in precopular position (i.e., the opposite of the situation in (1a))

6 the two major constituents of the specificational pseudocleft in (1b) can in fact change places, as in

(4a,b) — on a par with specificational copular sentences of the type in (5)

6 such word-order alternation impossible in predicational ‘pseudoclefts’, unless the AP predicate is

degree-modified — and in this respect, (1a) behaves exactly like other garden-variety copular

sentences with AP predicates (see (2b) and (2bN), and compare these to (3b,bN))

(2) a. what Johni is important to him i

b. *important to him i is what Johni is

bN. most important to him i is what Johni is

(3) a. getting there on time is important to him

b. *important to him is getting there on time

bN. most important to him is getting there on time

(4) a. what Johni is is important to himselfi

b. important to himselfi is what Johni is

(5) a. the best candidate is John

b. John is the best candidate

• that predicational ‘pseudoclefts’ behave like predicational copular sentences in general, and that

specificational pseudoclefts pattern with specificational copular sentences in general, is also apparent

from the distribution of the copula in non-finite contexts

(6) a. I consider what John i is (to be) important to him i

b. I consider getting there on time (to be) important to him

Marcel den Dikken — A cleft palette 2

(7) a. I consider what John i is *(to be) important to himselfi

aN. I consider important to himselfi ?(to be) what Johni is

b. I consider the best candidate *(to be) John

bN. I consider John (to be) the best candidate

6 the parallel between (7a,aN) (see Williams 1983; for some remarks on speaker variation, see Den

Dikken 2005:§3.4) and (7b,bN) is strongly suggestive of a parallel syntactic derivation for specificat-

ional pseudoclefts of the type in (1b) and inverse specificational copular sentences of the type in (5a)

(8) a. [SC [Subject important to himself/John] [Predicate what John is/the best candidate]]

b. [IP [Predicate what John is/the best candidate]k [BE [SC [Subject important to himself/John] tk]]]

1.2 It-clefts: Predicational vs specificational

(9) it was an interesting meeting that I went to last night

a. PREDICATIONAL — ‘the meeting I went to last night was interesting’

b. SPECIFICATIONAL — ‘I went to the following last night: an interesting meeting’

6 on the specificational reading the entire postcopular noun phrase is the focus of the it-cleft and

supplies new information

6 on the predicational reading only the attributive adjective interesting seems to supply new infor-

mation (and, concomitantly, only the adjective is focally stressed): meeting is discourse-old in (9b)

• Declerck (1988:158ff.) presents twenty characteristics that (9) on its predicational reading shares with

predicational sentences (see also Patten, this conf., for discussion of predicational clefts); I will turn

to one of these characteristics in section 2, below: the distribution of the copula in non-finite contexts

6 for now, what is perhaps most useful to note is that predicational and specificational it-clefts are

genuinely different syntactic creatures — as is perhaps particularly evident from the fact that the two

resist being conjoined (Declerck 1988:161)

(10) a. SPECIFICATIONAL & SPECIFICATIONAL

it was a BOOK that John gave me and a BIKE that Mary gave me

b. PREDICATIONAL & PREDICATIONAL

it was an IMPORTANT meeting that I went to and an INTERESTING subject that they discussed

c. ??PREDICATIONAL & SPECIFICATIONAL

??it was an IMPORTANT meeting that I went to and JOHN who was presenting a talk at it

d. ??SPECIFICATIONAL & PREDICATIONAL

??it was JOHN who led the debate and an INTERESTING subject that they discussed

6 as Declerck notes correctly, the problem with (10c,d) is presumably of the same type as that with (11)

(11) a. *John is the tall one and also fat

b. *John is both the bank robber and very charming

1.2 A more fine-grained typology: Subspecies of specificational cleft sentences

1.2.1 Specificational pseudoclefts

• based on Declerck (1988) (with some terminology changes) [new information underlined]


1 On NPI-licensing in Wolof clefts, see Torrence (this conf.). In English it-clefts, NPI-connectivity effects are never possible.

(12) types of SPECIFICATIONAL pseudoclefts

a. CONTRASTIVE or STRESSED-FOCUS PSEUDOCLEFTS

what broke the camel’s back? — what broke the camel’s back was a straw

— a straw was what broke the camel’s back

b. CONTINUOUS-TOPIC PSEUDOCLEFTS

do you know Mary’s book? — yes, in fact Mary’s book is what got me interested in clefts

c. DISCONTINUOUS or ALL-NEW or BROAD-FOCUS PSEUDOCLEFTS

those apples are good, aren’t they? — so they are! what keeps me from eating all of them is

that mother would be furious if I left none for the others

Declerck 1988:224 (12a) (12b) (12c)

new vs old

information

VALUE new; wh-clause old VALUE old; wh-clause

new but presented as old

both VALUE and wh-clause

new

accentuation

pattern

VALUE: heavy; wh-clause:

weak

VALUE: weak; wh-clause:

normal

both VALUE and wh-clause

normal

contrast focus strongly not strongly not strongly

word order both orders grammatical wh-clause postcopular both orders grammatical

• Den Dikken, Meinunger & Wilder (2000): two types of contrastive pseudoclefts (also Hedberg, this conf.)

(13) a. ‘TYPE A’ CONTRASTIVE PSEUDOCLEFTS

– full–IP VALUE (subject to optional ellipsis)

– wh-clause necessarily precopular

– NPI connectivity possible1

what nobody bought was [IP (nobody bought) any wine]

*[IP (nobody bought) any wine] was what nobody bought

b. ‘TYPE B’ CONTRASTIVE PSEUDOCLEFTS

– XP VALUE (where X…I)

– wh-clause variably placed

– NPI connectivity impossible

what nobody bought was [NP wine]

[NP wine] was what nobody bought

6 Den Dikken, Meinunger & Wilder (2000) argue that ‘Type A’ contrastive pseudoclefts have the same

syntax as question-answer pairs, and analyse the wh-clause of ‘Type A’ pseudoclefts as ‘self-answer-

ing questions’ with the wh-question placed in the topic position of a TO PIC-COMMENT STRUCTURE

whose Top–head is occupied by the copula (whence *what nobody bought may have been (nobody

bought) any wine — to be contrasted with what nobody bought may have been wine)

(14) [TopP [Topic what nobody bought] [Top=BE [Comment=IP (nobody bought) any wine]]]

6 for ‘Type B’ contrastive pseudoclefts, an analysis wherein the wh-clause is treated as a free relative

serving as the underlying predicate of a SMALL CLAUSE and ending up in clause-initial position via

PREDICATE INVERSION (à la (8b)) is more likely to be correct


2 An example of a ‘thetic cleft’ is the following Swedish example (from Huber 2002:179), uttered in a context in whichsomeone just called Anna but got Anna’s husband on the phone: det var hennes MAN som svarade ‘it was her husband REL answered’.Huber calls these ‘thetic’ in an effort to assimilate such it-clefts to the ‘thetic judgements’ of Brentano (see Kuroda 1972); see alsoSasse’s (1987) ‘mise en relief’ constructions in French, also mentioned by Huber (2002:178).

3 Erades (1962) makes this observation as well, though he does not state it in categorical terms. Prince (1978:899) pointsout that of the 50+ examples of informative-presupposition it-clefts with a nominal focus that she had gathered, every one of theminvolved a subject — ‘While I certainly am not claiming that focused objects never occur in this construction, I do consider this animportant area for further investigation’. She does, however, point out that non-local subjects are readily focusable in these it-clefts;she built (16b) on an attested example with a highest-subject focus. Kayne (2008) gives a made-up example of a continuous-topicit-cleft with an object gap in the relative clause: do you know Mary? — yes, in fact it was Mary {who/?that} I learned linguistics fromin the first place. I will hence assume that continuous-topic it-clefts allow object gaps. See section 4 for further discussion.

6 for continuous-topic (Declerck’s ‘unaccented-anaphoric-focus’) pseudoclefts (type (12b)), which

categorically resist the WH<VALUE order (the wh-clause must be post-copular), an analysis exploiting

(8a) as the underlier is again available — but this time PREDICATE INVERSION is impossible because

the wh-clause and only the wh-clause harbours new information: Predicate Inversion constructions

are generally incompatible with an information-structural signature in which the inverted predicate

represents new information (cf. what’s John’s job? — John’s the emcee / *the emcee is John)

1.2.2 Specificational it-clefts

• like specificational pseudoclefts, specificational it-clefts come in at least three flavours (see Erades

1962, Prince 1978, Declerk 1988, Hedberg 2000, Huber 2002, Doetjes, Rebuschi & Rialland 2004)

(15) a. CONTRASTIVE or STRESSED-FOCUS IT-CLEFTS

what broke the camel’s back? — it was a straw that broke the camel’s back

b. CONTINUOUS-TOPIC IT-CLEFTS

do you know Mary’s book? — yes, in fact it was Mary’s book that got me interested in clefts

c. DISCONTINUOUS or ALL-NEW or BROAD-FOCUS IT-CLEFTS

those apples are good, aren’t they? — so they are! it’s the fact that mother would be furious

if I left none for the others that keeps me from eating all of them

6 Huber (2002) recognises some additional subtypes of it-clefts, but some of these do not seem to be

universal (his ‘thetic clefts’ are found in Swedish but, as he notes, not in German2), and others (in

particular, his ‘clefts with multiple foci’ and ‘I(ntonation)-clefts’) will likely reduce to type (15a);

I will ignore these here

• as in the case of contrastive specificational pseudoclefts, the VALUE/FOCUS of contrastive it-clefts

must be focally stressed, and the relative clause typically has a flat intonation

• in continuous-topic it-clefts, by contrast, the relative clause conveys new information and hence

attracts prosodic prominence

• Erades (1962)/Prince (1978) and Kayne (1981) note two important syntactic properties of continu-

ous-topic it-clefts (Prince’s ‘informative-presupposition it-clefts’) [neither shared by all speakers]

(i) ‘Informative-presupposition it-clefts are formally and unambiguously identifiable. First, unlike

stressed-focus it-clefts, they have normally (vs. weakly) stressed that-clauses.[fn. omitted] Second,

they generally have a short and anaphoric focus, which, in my data, is either a (subject) NP or an

adverbial, generally of time, place, or reason ... Third, in such sentences, that/wh- is not deletable.’

(Prince 1978:898; my underscore)3


4 Huber (2002:176) points out for German continuous-topic it-clefts that a pronominal VALUE must be in sentence-initialposition — but for German there does not appear to be any difference between continuous-topic and contrastive it-clefts in this regard,judging from Huber’s (2002:78) observation that cleft pronouns should as a general rule be placed in sentence-initial position inGerman clefts (i.e., Huber rejects *es ist er, der kommt ‘it is he who comes’). For Dutch facts are clearly different. See also below.

(16) a. I want you to agree with me, no matter who it is {that/i} you think is right

b. the leaders of the militant homophile movement in America generally have been young

people; it was they {who/*i} everyone knows fought back during a violent police raid on

a Greenwich Village bar in 1969

(ii) Kayne (1981:§3.3.3): that cannot be used in highest-subject it-clefts of the continuous-topic (or

informative-presupposition) type if the VALUE is [+human]

(17) A: do you know Mary’s book?

B: yes, in fact it was Mary’s book {which/that} got me interested in clefts

(18) A: do you know Mary?

B: yes, in fact it was Mary {who/?*that} got me interested in clefts


B: yes, in fact it was Mary {who/that} everybody knows got me interested in clefts

6 to these generalisations about English continuous-topic it-clefts, we can add a th ird with reference

to Dutch

(iii) while in Dutch contrastive or stressed-focus it-clefts in root contexts the VALUE/FOCUS can be either

be in postcopular position (as in English) or in sentence-initial position, in continuous-topic it-clefts

the VALUE must always be in sentence-initial position4

[judging from Gryllia & Lekakou 2007, this is roughly the reverse of the situation in Cypriot Greek]

(20) A: wie heeft je interesse voor clefts gewekt? (Dutch)

‘who got you interested in clefts?’

B1: het was zij die m’n interesse voor clefts wekte

‘it was SHE who got me interested in clefts’

B2: zij was het die m’n interesse voor clefts wekte

(21) A: ken je haar?

‘do you know her?’

B1: ?*ja, het was zij die m’n interesse voor clefts wekte

‘yes, it was she who got me interested in clefts’

B2: ja, zij was het die m’n interesse voor clefts wekte

6 this pattern does not manifest itself in English (or Swedish, for that matter; see Huber 2002; nor does

it seem to be replicated in this exact form in German: see fn. 4, above)

(22) A: do you know her?

B1: yes, in fact it was she who got me interested in clefts

B2: *yes, in fact she was it who got me interested in clefts

vs yes, in fact she was the one who got me interested in clefts

6 in English, the it of it-clefts is always in the structural subject position — a point that I will return

to in the discussion of the status of it in section 2, below


5 I will have very little to say in this paper about the third type of it-clefts, the DISCONTINUOUS or ALL-NEW type in (15c).

• three syntactic generalisations thus emerge about continuous-topic it-clefts, setting them apart from

other it-clefts

(23) a. in CONTINUOUS-TOPIC IT-CLEFTS, the relative clause cannot have a null left edge: a wh-word

or that must be used, i is impossible (Erades 1962, Prince 1978)

b. in CONTINUOUS-TOPIC IT-CLEFTS with a [+human] highest-subject gap, the relative clause is

introduced by who, not by that (Kayne 1981)

c. in CONTINUOUS-TOPIC IT-CLEFTS in Dutch root contexts, the VALUE must be in sentence-

initial position (hence het ‘it’ cannot be in that position, and must instead be postcopular)

Q do these generalisations add up to a coherent overall picture of continuous-topic it-clefts (as distinct

from contrastive/stressed-focus it-clefts)?5

6 to be (partially) revisited in section 4, where continuous-topic it-clefts will be at centre-stage

2 On the it and be of it-clefts

• Dutch it-clefts and agreement

(24) a. het was ZIJN VROUW die... het waren ZIJN KINDEREN die...

it was his wife who... it were his children who...

b. ZIJN VROUW was het die... ZIJN KINDEREN waren het die...

his wife was it who... his children were it who...

(25) a. dat het ZIJN VROUW was die... dat het ZIJN KINDEREN waren die...

that it his wife was who... that it his children were who...

b. dat ZIJN VROUW het was die... dat ZIJN KINDEREN het waren die...

that his wife it was who... that his children it were who...

6 the thing to note about these examples, besides the freedom in word order (to which I will return

momentarily), is that no matter where the VALUE is placed, it always controls the agreement with the

finite copula; het never agrees with the copula in Dutch it-clefts (see fn. 9 for some relevant remarks)

6 this is the first respect (of several to come) in which it-clefts behave like specificational copular

sentences with a nominal predicate (see (26)–(27)) — not surprisingly, in light of the fact that they

are, after all, specificational copular sentences

(26) a. het grootste probleem zijn/*is de kinderen

the biggest problem are/is the children

b. de kinderen zijn/*is het grootste probleem

the children are/is the biggest problem

(27) a. dat het grootste probleem de kinderen zijn/*is

that the biggest problem the children are/is

b. dat de kinderen het grootste probleem zijn/*is

that the children the biggest problem are/is

• Dutch it-clefts and order


6 See Huber (2002:78) for German’s strong preference for b–order, even in the root, when the VALUE is pronominal. ForSwedish, Huber (2002:88) points out that in V2 root clauses a contrastively stressed pronominal VALUE may (somewhat marginally)be placed in sentence-initial position: (?)

HAN är det som kommer ‘he is it REL comes’. He does not comment on non-root contexts.

7 Huber’s (2002:79) observations regarding word order in German embedded it-clefts with the demonstrative das ‘that’ asthe VALUE go in the same direction: daß es das ist was mir Spaß macht ‘(lit.) that it that is what me pleasure makes, i.e. that it’s thatwhich gives me pleasure’ vs. ?daß das es ist was mir Spaß macht ‘that that it is what me pleasure makes’. But for personal pronounVALUES, Huber’s judgements show a pattern that is strikingly different from that seen in (my) Dutch: in Huber’s German, there isactually slightly more word-order flexibility in the non-root context than there is in the root context (i.e., quite the opposite of thesituation in (my) Dutch). I will have nothing insightful to say about the difference between Dutch and German.

8 In Dutch, the nominative forms of the third-person feminine singular and common-gender plural pronouns are identical,being differentiated by the agreement they trigger on the verb.

9 Note, though, that in root clauses ?het was ik die... ‘it was I who...’ and ?het was jij die... ‘it was you who...’, whilethemselves not brilliant, contrast sharply with present-tense *het ben ik die... ‘it am I who...’ and *het ben jij die... ‘it are you who...’— presumably because of a phi-feature agreement problem (which cannot, by the way, be avoided by having het control agreement:*het is ik/jij die... is impossible, too); uninverted ik ben het die... and jij bent het die... are perfect. The root of the agreement problemwith *het ben ik/jij die... must lie in the fact that the inverted predicate in it-clefts is pronominal: in inverse copular sentences witha full-nominal predicate, such as de beste kandidaat ben ik/jij ‘the best candidate is me/you’, the form ben ‘am/are(2SG-INV)’ is per-fectly fine. Apparently, though het never itself controls phi-feature agreement with the finite verb, it prevents the postcopular subjectfrom controlling phi-feature agreement if the postcopular subject is specified for first or second person — for reasons unclear to me.

6 as (24) and (25) show, when the focus is a full DP, both het<VALUE and VALUE<het orders are

possible, both in root and in non-root contexts, with a slight preference, in both contexts, for the

a–pattern6

6 when the VALUE is a pronoun, the two orders in (28) are still possible (though this time with a prefer-

ence for the b–order), but in the non-root environment in (29), the het<VALUE order is now unaccept-

able (Google statistics support this intuition robustly: while there are thousands of tokens of type

(29b), there are exactly six hits for the string dat het zij was die and exactly three for the string dat

het zij waren die (Nov 2008), and the hits returned for these strings all sound highly unnatural to me)7

(28) a. het was/waren ZIJ die...8

it was/were she/they who...

b. ZIJ was/waren het die...

she/they was/were it who...

(29) a. ?*dat het ZIJ was/waren die...

that it she/they was/were who...

b. dat ZIJ het was/waren die...

that she/they it was/were who...

6 this is not a peculiarity of third-person pronouns (whether singular or plural): first- and second-

person pronouns give the same picture — ?*dat het ik was die... ‘that it I was who...’ and ?*dat het

jij was die... ‘that it you were who...’ both sound poor and deliver very few hits on Google; dat ik het

was die... ‘that I it was who...’ and dat jij het was die... ‘that you it were who...’, by contrast, are

perfectly natural and common9

• the word-order freedom in root contexts in Dutch it-clefts is likely a simple reflex of the fact that

Dutch is a V2 language using the prefinite position in root clauses freely for any sentence constituent

— in other words, the alternation in (24a,b) and (28a,b) is likely an AN–alternation, featuring

AN–fronting into SpecCP in one of the two orders


10 The pattern observed in (28) vs (29) is also observable in another context in which AN–fronting to SpecCP is independentlyimpossible: yes/no–questions, and also wh-questions with a filled SpecCP. But though the pattern observed in (28) vs (29) is manifest,it is somewhat less homogeneous than it is in the root/non-root contrast — in particular, though (iia) with waren is impossible forme (on a par with the observations about non-root contexts in the main text), I find (iia) with singular was relatively passable.

(i) a. (waarom) was het ZIJN VROUW die...? (waarom) waren het ZIJN KINDEREN die...?why was it his wife who... why were it his children who...

b. (waarom) was ZIJN VROUW het die...? (waarom) waren ZIJN KINDEREN het die...?why was his wife it who... why were his children it who...

(continued...)

• since V2 and AN–fronting to SpecCP are impossible in all non-root contexts in Dutch, the word-order

freedom seen in (25a,b) cannot be analysed as an AN–alternation — instead, it likely reflects the kind

of word-order alternation seen in specificational nominal predication constructions more generally

(30) a. dat het probleem zijn vrouw was dat het probleem zijn kinderen waren

that the problem his wife was that the problem his children were

b. dat zijn vrouw het probleem was dat zijn kinderen het probleem waren

that his wife the problem was that his children the problem were

(31) a. ?*dat het probleem zij was/waren

that the problem she/they was/were

b. dat zij het probleem was/waren

that she/they the problem was/were

6 here again we find a difference in behaviour between full–DP and pronominal subjects: (31a) is much

worse than (30a)

• (30a) and (31a) are Predicate Inversion constructions, arguably derived via A–movement of the

predicate to SpecIP (see Moro 1997, Den Dikken 2006)

6 the fact that (29a) is significantly worse than (29b) suggests parallel behaviour to Predicate Inversion

constructions: (29a) is parallel to (31a), with both being much worse than their alternatives

6 this supports an approach to the underlying structure of it-clefts according to which it is structurally

represented as the PREDICATE of a small clause (cf. also Cheng & Downing, this conf., on Zulu clefts),

undergoing Predicate Inversion (i.e., A–movement to SpecIP) whenever it precedes the copula in an

environment in which such a word order could only result from movement to SpecIP (i.e., in non-root

contexts; in root contexts in Dutch, there is always the possibility of AN–fronting to SpecCP)

(32) a. [sc [Subject VALUE] [Predicate it]]

b. [IP iti ... [sc [Subject VALUE] [Predicate ti]]]

(in root and non-root contexts, EXCEPT if Subject = PRONOUN)

c. [CP iti ... [IP VALUEk ... [sc [Subject tk] [Predicate ti]]]]

(ONLY in root contexts; Subject unrestricted)

6 Predicate Inversion is apparently allowed in Dutch (for reasons that need not concern us; I will use

this as a diagnostic and will not present it as an explanandum here) when the small clause has a

full–DP subject but not (or at least, not very easily) when the small-clause subject is pronominal

6 familiar copular inversion constructions of the type in (31) behave on a par here with it-clefts, so it-

clefts are they way they ought to be if the derivation of their it<VALUE order (in contexts in which

AN–fronting of the predicate is unavailable) involves Predicate Inversion10


(ii) a. (waarom) {?(?)was/*waren} het ZIJ die...?why was/were it she/they who...

b. (waarom) was/waren ZIJ het die...?why was/were she/they it who...

(iii) a. (waarom) {??was/*waren} het probleem zij?why was/were the problem she/they

b. (waarom) was/waren zij het probleem?why was/were she/they the problem

While I do not have anything to say about the singular/plural distinction in (ii) and (iii), what matters for our purposes here is thatthere is once again an empirical parallel between it-clefts and standard copular inversion constructions — a parallel that furtherenhances the analytical similarity between the two construction types.

11 Though Declerck (1988) notes this himself, he nonetheless believes that predicational it-clefts ‘are essentially specification-al’ (except for the fact that the focused modifier is entirely predicational). It seems to me that treating the predicational it-cleft as asubspecies of specificational sentences would be a mistake.

12 Support for this conclusion comes from Ball’s (1978) observation (quoted by Hedberg, this conf.) that predicational cleftswith a plural VALUE take a plural pronoun/demonstrative rather than it (those are real eye glasses that Micky is wearing), whereasspecificational clefts with a plural VALUE stick with it. Many of the examples of ‘predicational it-clefts’ in Declerck (1988) involvepostcopular noun phrases that are predicable of non-humans, so it as the subject of these predicates is perfectly feasible for thosecases. But there are also predicational it-clefts that have postcopular predicates that should be predicated of humans, like it’s noRELIABLE man that you have hired, but a crook! (p. 160). The use of referential it as the subject here raises some questions.

• English it-clefts and the distribution of the copula (Declerck 1988:171)

(33) a. I consider John (to be) his best friend

b. I consider his best friend *(to be) John

(34) a. I consider it (to be) an INTERESTING subject that they are discussing tonight

b. I consider it *(to be) JOHN who is his best friend

6 the predicational it-cleft in (34a) does not force the presence of to be, which makes it behave on a par

with canonical, non-inverted, predicational copular sentences of the type in (33a)11

6 so it in (34a) is the subject of the predicate an INTERESTING subject, just as John in (33a) is the subject

of the predicate his best friend

6 that makes the it of (34a) a referential pronoun12

6 like the it of predicative it-clefts such as (34a), the proleptic it of clausal extraposition constructions

of the type in (35a) likewise does not force to be to be present in the complement of consider (35b)

(35) a. it is clear/important/unfortunate that S

b. I consider it (to be) clear/important/unfortunate that S

6 arguably, the proleptic it of clausal extraposition constructions is an argument — see esp. Bennis

(1986) on the idea that English it is never truly expletive, and serves as an argument in clausal extra-

position constructions

6 the fact that this it can serve as a controller of PRO is perhaps Bennis’s strongest argument to this

effect (Bennis’s examples in this context are from Dutch, but the point carries over to English, too)

(36) iti finally became clear, after PRO i having been explained ten times, that the earth is round

6 constructing similar control examples for predicational it-clefts is not an easy matter — but it seems

to me that (37) makes the point that it is an argumental pronoun in these constructions as well


13 Reeve (2007 and this conf.) presents (ia,b) as grammatical, contradicting the text claim. I cannot weigh in on the Englishjudgements here, but note that Dutch genuinely resists control by het in specificational it-clefts. I will take such control to begenerally impossible, leaving Reeve’s examples aside. While one could in principle take (38) to indicate that the it of contrastive/stressed-focus it-clefts is an ‘expletive’, Huber (2002:140) shows that in German it really does not behave as such. As is well known,‘truly expletive’ es in German (as found e.g. in (ii)) occurs only in SpecCP of root clauses: (iib) is ungrammatical. By contrast, thees of German it-clefts does occur in non-SpecCP positions, as (iii) shows (see also Reeve). So the es of it-clefts is not an ‘expletive’.

(i) a. iti was John who Bill spent all his time with [despite PROi being Mary the day before] (Reeve 2007,b. iti was the furniture that annoyed John on Sunday [despite PROi being the décor the day before] this conf.)

(ii) a. es spielen Kinder auf der Straße b. Kinder spielen (*es) auf der Straßeit play children on the street children play it on the street

(iii) a. es ist Peter, der kommt b. Peter ist *(es), der kommtit is Peter who comes Peter is it who comes

14 If copula contraction is a diagnostic for specificationality (as some have claimed; see e.g. Kaisse 1979:708–9), then it isinteresting to note that it allows contraction in ALL it-clefts (except in it is I (who/that S), perhaps). I suspect that, since it would behard to deny (in light of the facts discussed above) that specificational it-clefts exist and involve a Predicate Inversion derivation,the distribution of copula contraction is probably not a reliable diagnostic for specificationality.

(37) a. iti was an INTERESTING meeting I went to last night, despite PRO i being poorly organised

b. iti was, despite PRO i being poorly organised, an INTERESTING meeting I went to last night

6 so the fact that the it of (35) and the it of predicative it-clefts behave on a par with respect to the

distribution of the copula under consider and with respect to control confirms the conclusion that the

it of predicative it-clefts is argumental — a referential pronoun (recall fn. 12, and Hedberg, this conf.)

6 by contrast, the contrastive/stressed-focus it-cleft in (34b) patterns with the inverse copular sentence

in (33b) in forcing to be to be present, suggesting that this it-cleft is derived similarly to the way

(33b) is derived — via Predicate Inversion

6 what may further underpin the conclusion that the it of contrastive/stressed-focus it-clefts is not an

argument but instead an inverted predicate is the fact that, as Huber (2002:136) points out, this it

cannot readily control PRO: (38)13

6 in this respect, the it of contrastive/stressed-focus it-clefts behaves differently from the proleptic it

of clausal extraposition constructions and also, judging from (37), from the it of predicational it-clefts

6 on the other hand, both the control facts and the copula distribution facts forge a direct link between

contrastive/stressed-focus it-clefts and other specificational sentences with predicate–be–subject

order — see (39) regarding control in the latter (from Huber 2002)

6 so it in (34b) must be the predicate (hence cannot control: control is the prerogative of arguments),

inverting with its subject (the VALUE John)14

6 the conclusion that it in English specificational it-clefts is an underlying predicate is of course in line

with the conclusion drawn previously on the basis of the Dutch word-order facts — see (32)

(38) ???iti is Peter who is coming without PRO i being a nice man (Huber 2002:136)

(39) a. ???the murdereri is the butler without PRO i being a bad guy (Huber 2002:127)

b. ???[who murdered John]i was the butler without PRO i being a bad guy (Huber 2002:134)

6 it bears demonstrating that it is not just contastive/stressed-focus it-clefts that respond positively to

the diagnostics for specificationality and Predicate Inversion — continuous-topic it-clefts likewise

force the presence of to be under verbs like consider, as (40) shows


B: yes, in fact I consider it *(to be) Mary who got me interested in clefts

6 as a family, therefore, specificational it-clefts behave like other specificational copular sentences;

their derivation arguably involves Predicate Inversion


15 It is worth noting in passing that it is impossible to derive (43b) from an input similar to the structure assigned to predi-cational it-clefts, with it serving as a referential pronoun. Not only does this make no semantic sense, it also fails syntactically: theobligatoriness of to be under consider makes it clear that (43b) really is specificational (what do you consider it *(to be) that...).

• there is one important respect in which specificational it-clefts behave very differently from other

specificational copular sentences — wh-extraction15

(41) a. I think the most important item on the reading list is this book

b. *which book do you think the most important item on the reading list is t?

(42) a. I think what every linguistics student should read is this book

b. *which book do you think what every linguistics student should read is t?

(43) a. I think it is this book that every linguistics student should read

b. which book do you think it is t that every linguistics student should read?

6 Den Dikken (2006:125) claims that the ban on (41b)/(42b) is reducible to the ‘frozenness’ of foci:

‘A constituent that ends up in a syntactic configuration that leads it to be interpreted as a focus will

inevitably be interpreted as the focus of the clause that it is in, and will literally be frozen in place’

6 developing this account somewhat further, I believe there is a way of rendering (43b) compatible

with the general picture regarding extraction of the focus in specificational copular sentences:

– in all of the a–examples in (41)–(43), focus is assigned in postcopular position

– but only in (41a) and (42a) does the postcopular focus surface to the right of the topic: in the

it-cleft in (43a), the focus surfaces to the left of the topic (i.e., the material in the that-clause)

[the it of it-clefts evidently is not itself the topic of the it-cleft; so in it-clefts, unlike in other

inverse specificational copular sentences, the raised predicate does not represent the topic]

– so in (41b) and (42b) the focus is extracted across the topic , reversing the information-

structural pattern assigned as a result of Predicate Inversion

– in (43b), by contrast, extraction of the focus does NOT lead to a reversal of the information-

structural pattern of the Predicate Inversion construction: the focus still precedes the topic

Y extraction of the focus of a specificational copular sentence is grammatical so long as the

information-structural pattern (FOC<TOP or TOP<FOC) of the input to extraction is kept intact

3 On the relative clause of it-clefts

3.1 Parallels between the that/wh-clause of it-clefts and restrictive relative clauses

• of the three major constituents of the it-cleft, the one that is without a doubt the most elusive is the

that/wh-clause, called ‘relative clause’ in the header of this section

6 there are proper reasons for calling the that/wh-clause of it-clefts a relative clause, because there are

a number of non-trivial parallels between the that/wh-clause of it-clefts and restrictive relative clauses

(i) in many cases (for exceptions, see section 3.2, below), the that/wh-clause of it-clefts can be

introduced by either a wh-element or that, as is also the case in the typical (non-subject)

restrictive relative clause

(ii) in Dutch and German it-clefts with the demonstrative pronoun (D dat, G das ‘that’) as the

VALUE, just as in restrictive relative clauses headed by dat/das, the operator in the relative

clause is wat/was ‘what’ rather than dat/das ‘d-pronoun’ (daß es DAS ist was/*das mir Spaß

macht ‘that it that is what/d-pronoun me pleasure makes, i.e., that it is that which gives me

pleasure’ — Huber 2002:79; cf. das was/*das mir Spaß macht ‘that which gives me

pleasure’), whereas in it-clefts and headed relatives with other pronominal VALUEs/heads the

relative clause is always introduced by a d-pronoun


16 These six points are identified in the list below by attributions to Declerck’s book.

[in this context, note also the Dutch proverbial it-clefts het zijn niet allen koks die lange

messen dragen ‘it isn’t all cooks who(=d-pronoun) carry long knives’ and het is niet alles

goud wat er blinkt ‘it isn’t all gold what there glitters, i.e., all that glitters isn’t gold’]

(iii) in (Cypriot) Greek clefts (see Gryllia & Lekakou 2007), ‘we get the same pattern of resump-

tion in the pu-clause [of an it-cleft] as we do in [restrictive and free] relative clauses’:

impossible with accusative arguments, obligatory with non-arguments

6 yet besides these non-trivial parallels between garden-variety restrictive relatives and the that/wh-

clause of it-clefts, there are a significant number of ways in which the two diverge quite strikingly

3.2 Differences between the that/wh-clause of it-clefts and restrictive relative clauses

• integrating observations by many others before him, Declerck (1988:152, fn. 4) shows that the

relative clause of a specificational it-cleft ‘does not behave like a genuine restrictive relative clause

in several respects’

6 he mentions six respects in which the relative clause of it-clefts differs from a garden-variety restric-

tive relative clause16

6 I have added a few of my own, and have ordered the list in a way that makes good presentational

sense, with properties of the relative clause as a constituent listed before properties of the relative

operator and/or the VALUE constituent

(I) PROPERTIES OF THE RELATIVE CLAUSE AS A CONSTITUENT

(a) the relative clause of an it-cleft must always be clause-final, whereas a garden-variety

restrictive relative clause does not have to be ‘extraposed’ — this is particularly evident in

head-final languages such as Dutch, where the relative clause of an it-cleft can ONLY be

placed to the right of the finite copula regardless of whether the focus precedes or follows

het ‘it’; regular restrictive relative clauses, by contrast, do not need to ‘extrapose’

(44) a. dat het ZIJN ZOON <*die zojuist belde> was <Tdie zojuist belde>

that it his son who just called was who just called

aN. dat ZIJN ZOON <*die zojuist belde> het <*die zojuist belde> was <Tdie zojuist belde>

that his son who just called it who just called was who just called

b. dat ik met zijn zoon <die zojuist belde> heb gestudeerd <die zojuist belde>

that I with his son who just called have studied who just called

(b) the relative clause of an it-cleft cannot be an infinitival relative, whereas garden-variety

restrictive relative clauses can be (Declerck 1988:152)

(45) a. *it was the third man (who) to be arrested cf. it was the third man who was arrested

b. he was the third man (*who) to be arrested

(c) the relative clause of an it-cleft does not allow ‘relative clause reduction’, whereas a garden-

variety restrictive relative clause does (Declerck 1988:152)

(46) a. *it was JOHN working in the garden

b. John was the man working in the garden


(II) PROPERTIES OF THE RELATIVE OPERATOR AND/OR THE VALUE

(d) the relative clause of an it-cleft ‘can follow a uniquely referring “antecedent” such as a

personal pronoun (I, me) or a proper name not preceded by an article, whereas a restrictive

relative clause cannot’ (Declerck 1988:152)

(47) a. it is me that they invited

b. they were thinking of {a linguist/*me} that they could invite

c. it is John that they invited

d. *(the) John that they invited (was John Smith, not John Jones)

(e) the relative pronoun of an it-cleft ‘may sometimes be deleted in subject position ..., whereas

a subject relative pronoun can otherwise never be omitted (except in there-constructions)’

(Declerck 1988:152), in Standard English

(48) a. it was John saw it first

b. the police would like to talk to the person *(who/that) saw it first

(f) the relative clause of an it-cleft, when construed with a VALUE containing such, cannot be

introduced by as and must be introduced by that instead, whereas a garden-variety restrictive

relative clause construed with a head containing such must be introduced by as, not by that

(Declerck 1988:152)

(49) a. it is such a doll {that/*as} I’d like to have

b. this is such a doll {as/*that} I’d like to have

(g) the relative clause of an it-cleft, when construed with a non-locative PP VALUE, hence con-taining a non-locative PP–gap, is introduced by that, whereas a garden-variety restrictive

relative clause with a PP–gap cannot be

6 to this we can add the Dutch and German facts in (50N), with the German case (adapted from

Huber 2002:83) being particularly interesting: the only form usable to introduce the clause

is daß, the complementiser of finite clauses (as distinct from das, the d-word/demonstrative/

relative pronoun) — this is one of the most ‘dramatic’ ways in which the that/wh-clause of

it-clefts does not behave like a relative clause

(50) a. it was of Mary {that/*of whom} I was thinking

b. Bill was thinking of the woman {of whom/*that} Bob was thinking

(50N) a. het was aan Marie {dat/*aan wie/*waaraan} ik dacht (Dutch)

it was on Marie that/on whom/where-on I thought

b. es war an Maria {??daß/*an der/*woran} ich dachte (German)

it was on M aria that-COMP/on d-word/where-on I thought

(h) the relative clause of an it-cleft, when construed with a predicate nominal VALUE, can be

introduced by that, whereas a garden-variety relative clause construed with a predicate

nominal can only be introduced by which, not that or who (Declerck 1988:152)

6 to this we can add that (as Reeve 2007 points out) the relative clause of an it-cleft, when

construed with a predicate nominal VALUE, cannot in fact be introduced by which (cf. (52a))

(51) a. it is the president of the club that he has always wanted to be

b. Bill is the president of the club, {which/*that/*who} John has in fact always wanted to be

(52) a. *it is a doctor which I want to become, not a baker

b. it is a doctor that I want to become, not a baker


17 The more contentious case here is the which-example in (57b). The number of hits for (57b) (search: “what is/was itwhich”) is not huge (there are plenty of results that have “it” and “which” in separate clauses, such as what is it? which type is best?— but the number of clefts among the results is relatively small, and it may be significant that the bulk of the attested cases aresubject clefts: (i.i), below, is one of the rare cases of an object cleft with which), but there definitely are numerous examples of thistype, so it would not be right to exclude (57b) altogether. Below are some attested examples, culled from the internet with the aidof Google. For good measure, I have also included some attested examples of type (55b) (search: “who is/was it who”). Note thatexample in (iib) has a clearly non-specific/non-D-linked interpretation for who qua VALUE: so who does not have to be read as a D-linked element to make a relative clause introduced by who legitimate in an it-cleft.

(i) a. what is it which makes your product stand out from the rest?b. what is it which draws her into anger?c. what is it which prompts the generous, upon all occasions, and the mean upon many, to sacrifice their own

interests to the greater interests of others?d. what is it which makes say gold denser than aluminum?e. what is it which, if changed, makes a document essentially different?f. what is it which saves a soul?g. what is it which turns people away from God? what is it which tempts us to follow evil? what is it which blinds

us to Truth?h. what was it which led him to be such an effective double-agent?i. what was it which we set out to accomplish?j. what was it which got you interested in photography?

(ii) a. who is it who gets expensive jeans? is it Tutter or Malory?b. who is it who still thinks George W. Bush has the judgement, maturity, and emotional stability required to lead

the most powerful nation on earth?

18 Though (56b) may be much rarer than (56a) (just five hits on the web for “who the hell is it who”, and a few more thanfive for “who the hell was it who”; but several thousands of the type in (56a)), it is significant that it is not radically impossible forwho to be used as the relative operator in it-clefts with who the hell as the VALUE.

(i) the relative clause of an it-cleft, when construed with a(n aggressively) non-specific/non-D-

linked non-human VALUE, cannot be introduced by which, whereas a garden-variety restric-

tive relative clause with a non-specific/non-D-linked non-human head can be

6 this point requires some careful discussion

– note first that a specificity/D-linking effect of this sort seems manifest in it-clefts with

radically non-specific/non-D-linked anything at all or nothing at all as the VALUE, such as

the ones below: the a–sentences here are modelled on attested examples on the internet; and

while the a–pattern is not common (there are very few hits of this type on Google), it seems

to me significant that the b–pattern is not found at all (zero relevant hits for both “it is/was

anything at all which” and “it is/was nothing at all which”)

(53) a. it is very unlikely that it is anything at all that you are doing wrong

b. *it is very unlikely that it is anything at all which you are doing wrong

(54) a. it is nothing at all that you are doing wrong

b. *it is nothing at all which you are doing wrong

6 the effect can perhaps be observed most clearly for it-clefts with wh-the-hell as the VALUE

– while it-clefts with a [+WH] VALUE generally disprefer a wh-operator in the relative clause

(hence prefer that), examples of the type in (55b), with who, and (57b), with which, can be

found (though they tend to be rarer than corresponding examples with that)17

– by contrast, examples of the type in (58b), with which, are not found at all (and are judged

to be sharply worse than (58a) and also distinctly worse than (57b)), though it-clefts of the

type in (56b) are found18


(55) a. who is it that sends me anonymous E-mails all the time?

b. ?who is it who sends me anonymous E-mails all the time?

(56) a. who the hell is it that sends me anonymous E-mails all the time?

b. ?who the hell is it who sends me anonymous E-mails all the time?

(57) a. what is it that is bothering you? what is it that you want?

b. ?what is it which is bothering you? ?what is it which you want?

(58) a. what the hell is it that is bothering you? what the hell is it that you want?

b. *what the hell is it which is bothering you? *what the hell is it which you want?

6 so the D-linking/specificity effect seems to manifest itself only in it ... which clefts, not (or

at least not as robustly) in it ... who clefts

– this could plausibly be related to the fact that which is itself [+specific]/D-linked in wh-

phrases like which book, whereas who is not inherently marked for D-linking/specificity

– but the specificity of which does not normally ‘assert itself’ on the head of a relative clause

introduced by which — that is, in (59a) as well as (59b) we see an ambiguity between de

dicto (non-specific) and de re (specific) interpretations; the same is true for (60): though, to

my knowledge, no book specific exists that actually solves the mystery of it-clefts, (60b) is

perfectly felicitous, which, if which-relatives were to force a specific interpretation on the

relativised noun phrase, would be difficult to reconcile with our (or, at least, my) knowledge

of the world

(59) a. he wants to buy a car that is easy to park in small spaces

b. he wants to buy a car which is easy to park in small spaces

(60) a. he wants to find a book that solves the mystery of it-clefts

b. he wants to find a book which solves the mystery of it-clefts

6 so with respect to specificity/D-linking, the relative clause of it-clefts once again behaves

strikingly differently from garden-variety restrictive relative clauses

• having made an inventory (probably not exhaustive, but certainly representative) of properties of the

relative clause of it-clefts that make it similar to and different from garden-variety restrictive rela-

tives, let us discuss them now with an eye to gaining a better understanding of the nature of it-clefts

3.3 Obligatory ‘relative clause extraposition’

(44) a. dat het ZIJN ZOON <*die zojuist belde> was <Tdie zojuist belde>

that it his son who just called was who just called

aN. dat ZIJN ZOON <*die zojuist belde> het <*die zojuist belde> was <Tdie zojuist belde>

that his son who just called it who just called was who just called

b. dat ik met zijn zoon <die zojuist belde> heb gestudeerd <die zojuist belde>

that I with his son who just called have studied who just called

6 by analysing the wh-/that-clause as the complement of the copula (à la É. Kiss 1998), we could con-

nect (44a,aN) to obligatory extraposition of complement clauses to V in Dutch — BUT in English the

wh-/that-clause of it-clefts must also extrapose (Reeve 2007), while V–complement clauses need not

Q are there any contexts of obligatory relative clause extraposition (outside the realm of it-clefts)?

6 it is certainly not the case that a relative clause associated with the VALUE of an inverse specificat-

ional copular sentence is forced to extrapose


19 On the link between it-clefts and right-dislocation constructions, see also Frascarelli & Ramaglia (this conf.:§4.2).

20 Another empirical parallel between it-clefts and right-dislocation is that neither right-dislocated constituents nor the relativeclause of it-clefts can be stranded under sluicing: (i). (On VP–ellipsis in it-clefts and RC–extraposition, see Reeve 2007, this conf.)

(i) a. *I do believe some people like him, but I can’t figure out who likes him, President Bushb. *I remember what it was that Bill bought, but I don’t remember what it was that Bob bought

6 in the inverse predicate nominal constructions in (61a) and (62a), extraposition is perhaps preferred

but by no means obligatory

[and as a matter of fact, in the ‘canonical’ predicate nominal constructions in (61/2b), which also

support a specificational reading, relative clause extraposition from the subject is extremely difficult]

(61) a. dat het probleem zijn zoon <(?)die zojuist belde> is <Tdie zojuist belde>

that the problem his son who just called is who just called

b. dat zijn zoon <die zojuist belde> het probleem is <?*die zojuist belde>

that his son who just called the problem is who just called

(62) a. dat wat op tafel ligt het boek <(?)dat Jan me gegeven heeft> is <Tdat Jan me gegeven heeft>

that what on table lies the book that Jan me given has is that Jan me given has

b. dat het boek <Tdat Jan me gegeven heeft> wat op tafel ligt is <*dat Jan me gegeven heeft>

that the book that Jan me given has what on table lies is that Jan me given has

6 so the relative clause of an it-cleft does not behave like a garden-variety restrictive relative clause

associated to the subject of a predicate nominal construction — and more specifically, it does not

even behave like a garden-variety restrictive relative clause associated to the VALUE of an inverse

specificational copular sentence

• one context in which extraposition of something that looks like a relative clause is indeed obligatory

is RIGHT DISLOCATION, illustrated by English (63a–d) and Dutch (64a–d)19

(63) a. it was an explosion, what Bill heard

b. it’s apples, what I have enough of

c. it was Alice, the one who just had the baby

d. it was a model train, what I bought

dN. it was a model train {that/which} I bought

(64) a. dat het een explosie <*wat Bill gehoord heeft> was, <Twat Bill gehoord heeft>

b. dat het appels <*waar ik genoeg van heb> zijn, <Twaar ik genoeg van heb>

c. dat het Alice <*degene die een baby heeft gekregen> was, <Tdegene die zojuist een baby

heeft gekregen>

d. dat het een modeltrein <*wat ik gekocht heb> was, <Twat ik gekocht heb>

dN. dat het een modeltrein <*die ik gekocht heb> was <Tdie ik gekocht heb>

6 the examples in (63) are from Gundel (1977), who takes them to be the source for it-clefts; the

examples in (64) are their Dutch renditions

6 Gundel’s (1977) idea that the relative clause in it-clefts is in the same structural position as right-

dislocated constituents, while at first contradicted by the absence in clefts of the prosody typical of

right-dislocation, seems to be supported by the fact that the relative clause of clefts must indeed be

radically final, like right-dislocated constituents20

Q what is the analysis of right dislocation, and how does it shed light on the syntax of it-clefts?


21 Though Koster does in fact assume that the relative clause in (66a) is construed with the full DP een vrouw, nothing in hisproposal prevents a modification of the account wherein the relative clause is construed with just the NP vrouw: all that would needto be changed for this to materialise is the categorial status of the specifier of the ‘:P’ in (66a) (NP, excluding een, rather than DP).Since (66) involves a restrictive relative clause, and since restrictive relative clauses arguably associate (under direct construal) withNP rather than DP, I would favour the modified approach rather than the one Koster proposes. But nothing hinges on this matter here.

22 Koster’s ‘:P’ analysis of right dislocation and its extension to it-clefts on the next page are generally presented here as aconvenient outlook on the syntax of dislocation phenomena. Little of essence hinges on the adoption of ‘:P’; but see p. 20, re: (i).

6 Koster (2000) (see also Rijkhoek 1998, De Vries 1999) has proposed an ‘asyndetic coordination’

approach to right-dislocation (and, more generally, rightward extraposition) phenomena, with the

right-peripheral constituent introduced as the complement of a null conjunction (annotated as ‘:’) that

heads a phrase whose specifier is the portion of the sentence that the right-peripheral constituent

combines with

6 Koster introduces his approach to extraposition phenomena by first addressing the structure of

‘equatives’ of the type in (65), which in Dutch allow the specifying phrase (een gouden iglo ‘a golden

igloo’) to surface in a variety of surface positions

(65) a. Jan heeft iets moois, een gouden iglo, gebouwd

Jan has something beautiful a golden igloo built

[:P [DP iets moois] [: [een gouden iglo]]

b. Jan heeft iets moois gebouwd, een gouden iglo

Jan has something beautiful built a golden igloo

[:P [AgrOP iets moois gebouwd] [: [een gouden iglo]]

c. iets moois heeft Jan gebouwd, een gouden iglo

something beautiful has Jan built a golden igloo

[:P [CP iets moois heeft Jan gebouwd] [: [een gouden iglo]]

6 for relative clauses, extraposed as well as in situ , Koster proposes a parallel account21

(66) a. Jan heeft een vrouw die alles wist ontmoet

Jan has a woman who everything knows met

[:P [DP een vrouw] [: [die alles wist]]

b. Jan heeft een vrouw ontmoet die alles wist

Jan has a woman met who everything knew

[:P [AgrOP een vrouw ontmoet] [: [die alles wist]]

c. een vrouw heeft Jan ontmoet die alles wist

a woman has Jan met who everything knew

[:P [CP een vrouw heeft Jan ontmoet] [: [die alles wist]]

6 assume that (66b,c) represent the right approach to relative clause extraposition

6 assume further that this approach carries over to right dislocation22

(67) a. dat ik hem <*Piet> haat, <TPiet>

that I him Piet hate Piet

*[:P [DP hem] [: [Piet]]

[:P [AgrOP hem haat] [: [Piet]]

b. dat het een modeltrein <*wat ik gekocht heb> was, <Twat ik gekocht heb> (= (64d))

that it a model train what I bought have was what I bought have

*[:P [DP een modeltrein] [: [wat ik gekocht heb]]

[:P [IP het een modeltrein was] [: [wat ik gekocht heb]]


6 ‘extraposition’ (i.e., a structure in which the ‘:P’ takes as its specifier something larger than the noun

phrase with which the right-dislocated phrase is associated) is forced in these contexts because,

apparently, a direct asyndetic specification relationship between a pronoun and a full DP or between

a DP and a free relative (introduced, in Dutch (67b), by wat ‘what’) cannot be established

6 though I have no immediate insights to offer into the question of why a direct asyndetic specification

relation between a pronoun and a full DP is impossible (67a), for the case in (67b) it seems to me sig-

nificant that the nature of the relative clause here is different from that of the relative clause in (66):

– in (66) the relative clause is introduced by a d-pronoun, die, which allows it to be directly

construed with the ‘head’

– in (67b) the relative clause is a free relative, introduced by wat ‘what’, which prevents direct

construal with any ‘head’: by their very nature, free relatives are ‘headless’ or ‘null-headed’,

which should be taken to mean that they cannot be directly construed with any noun phrase

6 the analysis of (67b) is now readily extended to cover it-clefts

(68) a. dat het een modeltrein <*die ik gekocht heb> was <Tdie ik gekocht heb> (= (64dN))

that it a model train that I bought have was that I bought have

*[:P [DP een modeltrein] [: [die ik gekocht heb]]

[:P [IP het een modeltrein was] [: [die ik gekocht heb]]

b. it was a model train {that/which} I bought (= (63dN))

*[:P [DP a model train ] [: [{that/which} I bought]]

[:P [IP it was a model train ] [: [{that/which} I bought]]

6 for Gundel (1977), an ad hoc deletion rule (‘variable head deletion’) is responsible for the change

from (63d) to (63dN): the variable head of a dislocated relativised noun phrase is deleted, as a result

of which we (somehow) lose the what and we get that or which instead (details left vague by Gundel)

6 to get from (63d) to (63dN), or from Dutch (64d) to (64dN), the crucial question is what are which, that

and Dutch die, and the clauses introduced by them, in the cleft examples in (63dN) and (64dN)

6 plainly, the clauses introduced by that, which, die in the it-clefts in (68) had better not be garden-

variety restrictive relative clauses — for otherwise they ought to be allowed to stay in situ (or, put

differently, they ought to be allowed to be directly construed with a DP, as in the starred structures

below (68a,b))

6 we have already come across independent indications that the apparent relative clauses of it-clefts

are different from regular restrictive relatives — recall (I), above: in particular, the fact that reduced

relatives and non-finite relatives are impossible (see (45) and (46)); note also the fact that the relative

clause of it-clefts can associate with pronouns and proper names (recall (IId), above)

6 in these latter respects, the relative clause of it-clefts is really more like a free relative, which likewise

cannot be reduced or non-finite (what to do was unclear is grammatical but does not instantiate a free

relative: this is an infinitival wh-question instead; the free-relative case *what to do was important

is impossible, in contrast to what they should/would do was important)

6 but typical free relatives are never introduced by that, which or Dutch die

6 so let us gauge the status of the relative operators in it-clefts, and the nature of the clause that they

introduce, in more detail: this is the topic of section 3.4

3.4 The nature of the relative clause, its left periphery, and its relation to the VALUE

• there is evidence that which in the ‘relative clause’ of it-clefts behaves differently from the way it

does in garden-variety restrictive relatives — that is, it seems to assert its specificity/D-linking in

ways that it systematically does not do when it is the relative operator in ordinary restrictive relatives


23 This concord relationship is certainly present between a right-dislocated constituent and its clause-internal associate —and consequently, there is no right-dislocation paraphrase of the it-cleft in (ia): the right-dislocated constituent is definite and specific,so it is unable to entertain a concord relationship with the aggressively non-D-linked/non-specific wh-the-hell.

(i) a. what the hell is it {that/*which} you bought?b. what the hell is it, the thing {that/which} you bought?bN. what the hell {are they/*is it}, the things {that/which} you bought?

While (ib) is perfectly grammatical, it forces a referential reading upon it, with what the hell construed as the predicate nominal andthe right-dislocated constituent associated with it (see (ibN), which on this analysis represents a straightforward number mismatchbetween the right dislocate and its associate in the it-case) rather than what the hell — so (ia) and (ib) are very different kinds ofsentences. Importantly, (ib) cannot be read the same way as (ia) is supposed to be read, with it as the pro-predicate and what the hellas the subject — put differently, (ib) cannot be cleft-like.

24 In fact which itself is not necessarily specific: de dicto readings are normally possible in relative clauses: see esp. (i).

(i) we are expecting a student tonight; the student we are expecting tonight will likely leave early

6 this is apparent from two important sources noted above:

(IIh) the relative clause of an it-cleft, when construed with a predicate nominal VALUE, can and

must be introduced by that, whereas a garden-variety relative clause construed with a

predicate nominal can only be introduced by which, not that or who (Declerck 1988:152)

(51) a. it is the president of the club that he has always wanted to be

b. Bill is the president of the club, {which/*that/*who} John has in fact always wanted to be

(52) a. *it is a doctor which I want to become, not a baker

b. ?it is a doctor that I want to become, not a baker

(IIi) the relative clause of an it-cleft, when construed with a(n aggressively) non-specific/non-D-

linked non-human VALUE, cannot be introduced by which, whereas a garden-variety restric-

tive relative clause with a non-specific/non-D-linked non-human head can be

(53) a. it is very unlikely that it is anything at all that you are doing wrong

b. *it is very unlikely that it is anything at all which you are doing wrong

(54) a. it is nothing at all that you are doing wrong

b. *it is nothing at all which you are doing wrong

(57) a. what is it that is bothering you? what is it that you want?

b. ?what is it which is bothering you? ?what is it which you want?

(58) a. what the hell is it that is bothering you? what the hell is it that you want?

b. *what the hell is it which is bothering you? *what the hell is it which you want?

6 this suggests that which is, in some sense, the head of the relativised constituent, in a CONCORD

relationship with the VALUE of the cleft23

6 but how can which be heading a phrase that does not behave like a projection of a wh-constituent?

6 and how does this exclude the emergence of that alongside which, and how does it manage to accom-

modate subject-relatives with which, which would have to have a null operator and a null C (in effect,

they would have to be subject contact relatives, which Standard English does not otherwise allow)?

• so rather than representing which itself as the head of the relative clause, we need a NULL head that

agrees in specificity (a) with which (even though heads of relative clauses do not normally ‘agree’

with which in specificity24) and (b) with the VALUE of the it-cleft — as in (69)


According to Rullmann & Beck (1998) and Beck & Rullmann (1999), (i) demonstrates the need for reconstruction of the head of arelative clause into the relative clause. With a student in the first clause in (i) understood de dicto, we cannot take the student we areexpecting tonight to mean ‘the unique student x such that we are expecting x tonight’ (for there is no particular student that isexpected); reconstruction of student below expecting in the relative clause will deliver the desired result. (Note, however, that Sharvithas argued against the reconstruction-of-head analysis proposed by Beck and Rullmann.)

(69) [:P [it’s VALUE–XP["SPEC]] [‘:’ [XP i ["SPEC] [CP {which i/Opi}["SPEC] [CN {that/i} [IP (...) ti (...)]]]]]]

6 assuming that which is specified as [+SPEC], and that the i head of the relativised XP must show

CONCORD for the feature ["SPEC] with both the operator in SpecCP and the VALUE–XP, we now

derive the fact that whenever which is chosen as the operator of the relative clause in it-clefts, the

VALUE–XP must be [+SPEC]

6 this correctly rules out (52a), (53b), (54b), and (58b): in all of these cases, the nature of the VALUE–

XP is such that it radically resists being in a CONCORD relationship with a [+SPEC] element

6 the versions of these examples featuring that instead of which are grammatical because, unlike which,

the null operator in the specifier of that is not inherently specified for specificity, hence may assume

any specification for ["SPEC] that is appropriate in the structural environment

NB CONCORD between which, i, and the VALUE supplies a powerful argument against approaches to it-

clefts that take it and the relative clause to form a ‘discontinuous constituent’ (esp. à la Percus 1997)

Q why must i, the head of the relativised XP in it-clefts, necessarily be in a CONCORD relationship with

the relative operator and the VALUE–XP?

6 this question should be teased apart into two subquestions, regarding:

(i) the CONCORD relation between i and the relative operator

(ii) the CONCORD relation between i and the VALUE–XP

re: (i) hypothesis: i must show CONCORD with the relative operator in order to be formally licensed

– i is not licensed on its own

– i heads a constituent that is the complement of ‘:’ (see (69), and recall the discussion of

extraposition), and ‘:’ cannot formally license i because it is itself featureless

[it is here that my adoption of Koster’s ‘:P’ analysis is somewhat significant; recall fn. 22]

– i is structurally too distant from the VALUE–XP (embedded within the specifier of ‘:P’) to

be able to be formally licensed by it

6 i depends on its formal licensing on the relative operator, with which it must entertain a

CONCORD relation in order for formal licensing to be in effect

re: (ii) note that, while predicate nominals do not need to agree with their subjects in phi-features (not even

when the subject is pronominal), right-dislocated constituents do obligatorily share their phi-features

with the pronouns they are associated with, under CONCORD, regardless of whether the pronoun is

in subject or object position

(70) dat {de kinderen/zij} ons grootste probleem zijn

that the children/they our biggest problem are

(71) a. dat zij me irriteren, {die kinderen/*ons grootste probleem}

that they me irritate those children/our biggest problem

b. dat ik hen haat, {die kinderen/*ons grootste probleem}

that I them hate those children/our biggest problem

6 in an ‘:P’ structure of the type in (72), then, the complement of ‘:’ and the associated constituent (em-

bedded) in the specifier position of ‘:P’ must entertain a CONCORD relation


(72) [:P [IP/AgrOP PRONOUN[+PL] ...] [: [DP die kinderen/*ons grootste probleem]]]

6 so the fact that the i head of the relativised XP in (69) must show CONCORD with the VALUE–XP

enhances the parallel between it-clefts and right-dislocation constructions, captured via Koster’s ‘:P’

6 to derive the CONCORD requirement in (ii) in it-clefts, I hypothesise further that, under the CONCORD

relationship between i and the VALUE–XP, i is content-licensed

6 thus, i depends for its licensing on both the relative operator and the VALUE–XP: the former takes

care of formal licensing, and the latter of content licensing

• note that we can straightforwardly capitalise on the CONCORD requirement that we have just put in

place to account for the fact that, when the VALUE–XP is a PP, the gap inside the relative clause must

also be a PP — in other words, the difference between (73a) and (73b) follows from CONCORD; and

of course the fact that (73c) is grammatical is accommodated as well

(73) a. it is [PP to MARY] [XP=PP i [CP Op=PP [that I’d like to speak tPP]]]

b. *it is [PP to MARY] [XP=DP i [CP Op=DP [that I’d like to speak to tDP]]]

c. it is [DP MARY] [XP=DP i [CP Op=DP [that I’d like to speak to tDP]]]

6 what I do not have an airtight account for at this time is the fact that (73a) also becomes ungrammati-

cal when the null operator is replaced with a PP-pied-piping whom (*it is to MARY to whom I’d like

to speak) — I suspect that this has to do with general restrictions on physical pied-piping in i-headed

relative clause constructions, but I do not have a formal explanation in place

Q how is the relativised XP in (69) different from a free relative (which cannot be used in an it-cleft)?

6 put differently: what makes which- or that-relatives the only grammatical choice in the it-cleft in

(74a) while what-relatives (free relatives) are the only grammatical choice in the pseudocleft in (74b)?

(74) a. it’s KIMCHI {which/that/*what} I’d like to have

b. {what/*which/*that} I’d like to have is KIMCHI

6 in the it-cleft in (74a), the relationship between the null-headed relative and (the container of) the

VALUE–XP is established via ASYN DE TIC SPECIFICATION (see (75a)), which, as we recall from (71),

is subject to a strict CONCORD requirement

6 this strict CONCORD relationship facilitates the licensing of i

(75) a. [:P [it’s VALUE–XP] [: [XP i [CP {which i/Opi/*whati} [CN C [IP (...) ti (...)]]]]]]

b. [RP VALUE–XP [RELATOR [XP œ/DEF [CP {whati/*Op i/*which i} [CN C [IP (...) ti (...)]]]]]]

6 in the pseudocleft in (74b), the relationship between the free relative and the VALUE–XP is estab-

lished via PREDICATION (see the RELATOR–phrase (= ‘small clause’) in (75b,bN)), which, as we recall

from (70), is not subject to strict CONCORD between the subject and the predicate

6 due to the lack of a strict CONCORD relationship that i cannot be licensed in pseudoclefts

6 since the head of the relativised noun phrase is not allowed to be radically empty, something that has

feature content must head the relativised noun phrase in a pseudocleft

6 that ‘something’ could either be what itself (if what is effectively an indefinite pronoun — the

‘external head’ of a free relative) or, more likely (if what occupies the SpecCP of the relative clause,

like other wh-operators do in English), a phonologically null but semantically contentful operator:

œ or DEF(inite) (cf. ‘maximisation’ effects in free relatives; this may be related to the ‘exhaustivity’

property of specificational pseudoclefts); cf. Percus (1997) (who extends this to clefts, wrongly so)


6 all that now remains to be said to capture (74) in full is that (a) i is for all intents and purposes a

‘regular’ external head of a restrictive relative, being relativised with which or the null operator,

whereas (b) œ/DEF is an ‘unusual’ external head of a restrictive relative, forcing what as the relative

operator — cf. Dutch alles wat/*dat ‘everything what’ and dat wat/*dat ‘that what’ (recall §3.1, (ii))

[cf. the classic idea that what = it+which, now more appropriately rephrased as what=œ/DEF+which]

6 I have no insights to offer as to why (75b) delivers what — the analysis of free relatives is certainly

beyond the scope of this paper

6 but (75a,b) at least provide a window on the key difference between two types of null-headed rela-

tives with respect to the form of the relative operator: when the external head is radically empty (i),

the relative operator is the one expected for a garden-variety restrictive relative; it is only when the

external head is phonologically null but semantically specified (as œ or DEF) that the form of the

relative operator is unusual

• whatever ultimately underlies the unusual realisation of the relative operator in free relatives with a

phonologically null œ/DEF head, the points to take away from the discussion of (74a) vs (74b) are the

following:

(i) both it-clefts and pseudoclefts feature a null-headed relative clause

(ii) in it-clefts, the null head is radically empty (i.e., devoid of both phonological and semantic

features), hence dependent for its licensing on a CONCORD relation with the relative operator

and the VALUE–XP

(iii) the CONCORD relation between the i head, the relative operator, and the VALUE–XP is

responsible for the specificity effects seen in it-clefts featuring which as the relative operator

(see (52)–(54) and (58), and (IIh,i), above), and also explains the fact that when the VALUE

is a PP, the gap in the relative clause must necessarily also be a PP (see (73) and (IIg), above)

(iv) the fact that a strict CONCORD relation is established in an ASYNDETIC SPECIFICATION struc-

ture but not in a PREDICATION structure (a ‘small clause’) is responsible for the unavailability

of the i-headed relative (the default option) in pseudoclefts

• with all of this in place, let me go back to the list of properties of it-clefts and see how many more

of them can be accounted for

6 let me start with a review of the remaining properties listed under (II):

(IId) the relative clause of an it-cleft ‘can follow a uniquely referring “antecedent” such as a

personal pronoun (I, me) or a proper name not preceded by an article, whereas a restrictive

relative clause cannot’ (Declerck 1988:152)

(47) a. it is me that they invited

b. they were thinking of {a linguist/*me} that they could invite

c. it is John that they invited

d. *(the) John that they invited (was John Smith, not John Jones)

6 this is now basically straightforward: the relative clause is not in fact construed directly with

the VALUE, but instead with a i-head that entertains a CONCORD relationship with (but is not

identical to) the VALUE

(IIe) the relative pronoun of an it-cleft ‘may sometimes be deleted in subject position ..., whereas

a subject relative pronoun can otherwise never be omitted (except in there-constructions)’

(Declerck 1988:152)

(48) a. it was John saw it first

b. the police would like to talk to the person *(who/that) saw it first


25 Note that it does not seem to be a universal that the relative clause of it-clefts is always finite: Sleeman (this conf.) showsthat, unlike English and French, Italian allows it-clefts with infinitival subject relatives:

(i) è stato Gianni a darmi la chiave (Italian)is been Gianni to give-me the key‘it was Gianni who gave me the key’

Sleeman relates the contrast between Italian, on the one hand, and French and English, on the other, to the claim that Italian it-cleftsalways have a contrastive focus (whereas it-clefts in French and English have an identificational focus unmarked for contrastiveness),and that infinitival subject relatives are licensed specifically by contrastive focus. As a general claim about infinitival (subject) rela-tives this latter claim seems false (I’m looking for a man to fix the sink plainly does not require contrastive focus on a man). Notealso that infinitival relatives remain ungrammatical in English/French it-clefts even when the VALUE is actually contrastive. So itseems to me that, while Sleeman is probably right to implicate contrastiveness as a key difference between Italian and English/Frenchit-clefts, contrastiveness cannot be the sole agent in adjudicating the distribution of finite and infinitival relative clauses in it-clefts.

6 this is something that does not follow immediately from the structure proposed but may be

related, perhaps, to the fact that the head of the relative clause in it-clefts is itself null; I will

need to think about this more (if only because the empirical fact of the matter is not perfectly

straightforward: (48a) is definitely not universally accepted by Standard English speakers)

(IIf) the relative clause of an it-cleft, when construed with a VALUE containing such, cannot be

introduced by as and must be introduced by that instead, whereas a garden-variety restrictive

relative clause construed with a head containing such must be introduced by as, not by that

(Declerck 1988:152)

(49) a. it is such a doll {that/*as} I’d like to have

b. this is such a doll {as/*that} I’d like to have

6 this is something that is arguably relatable to the null-headedness of the relative clause con-

struction in the it-cleft: one may hypothesise that as is only licensed as a relative operator

in relative clauses whose external heads include the word such; the relative clause of it-cleft

constructions, by contrast, is headed by i, which does not include such, so as is not licensed

6 next, let us review the two remaining points brought up under (I):

(Ib) the relative clause of an it-cleft cannot be an infinitival relative, whereas garden-variety

restrictive relative clauses can be (Declerck 1988:152)

(45) a. *it was the third man (who) to be arrested cf. it was the third man who was arrested

b. he was the third man (*who) to be arrested

(Ic) the relative clause of an it-cleft does not allow ‘relative clause reduction’, whereas a garden-

variety restrictive relative clause does (Declerck 1988:152)

(46) a. *it was JOHN working in the garden

b. John was the man working in the garden

6 it seems to me that these two observations are intimately related to the fact that a free relative is

likewise never non-finite/reduced

6 in the case of free relatives this may be thought to be related to the fact that their left periphery may

not be empty (i.e., they must feature an overt relative operator, what, who, etc.)

6 since the relative clause of it-clefts is not required to have something overt in its left periphery, such

a suggestion would fail to carry over to it-clefts

6 at this time, I do not fully understand why the relative clause of English it-clefts and free relatives

is always finite — this, then, is a topic I must leave for some other occasion25


26 I will not be able to address property (iii), the fact that Dutch continuous-topic it-clefts want the VALUE to precede het ‘it’not just in non-root contexts but in root clauses as well (see (21)). I have no profound insights to offer into this matter at this time.

4 Contrastive/stressed-focus it-clefts versus continuous-topic it-clefts

• in section 1, I pointed out that, like specificational pseudoclefts, specificational it-clefts come in at

least three flavours

(15) a. CONTRASTIVE or STRESSED-FOCUS IT-CLEFTS

what broke the camel’s back? — it was a straw that broke the camel’s back

b. CONTINUOUS-TOPIC IT-CLEFTS

do you know Mary’s book? — yes, in fact it was Mary’s book that got me interested in clefts

c. DISCONTINUOUS or ALL-NEW or BROAD-FOCUS IT-CLEFTS

those apples are good, aren’t they? — so they are! it’s the fact that mother would be furious

if I left none for the others that keeps me from eating all of them

6 while I will have nothing to say about (15c), I would like to close this paper by discussing two of the

three salient differences between (15a) and (15b) enumerated in section 1.2.2, repeated here:26

(i) ‘Informative-presupposition it-clefts are formally and unambiguously identifiable. First, unlike

stressed-focus it-clefts, they have normally (vs. weakly) stressed that-clauses.[fn. omitted] Second,

they generally have a short and anaphoric focus, which, in my data, is either a (subject) NP or an

adverbial, generally of time, place, or reason ... Third, in such sentences, that/wh- is not deletable.’

(Prince 1978:898; my underscore)

(16) a. I want you to agree with me, no matter who it is {that/i} you think is right

b. The leaders of the militant homophile movement in America generally have been young

people. It was they {who/*i} everyone knows fought back during a violent police raid on

a Greenwich Village bar in 1969

(ii) Kayne (1981:§3.3.3): that cannot be used in highest-subject it-clefts of the continuous-topic (or

informative-presupposition) type if the VALUE is [+human]

(17) A: do you know Mary’s book?

B: yes, in fact it was Mary’s book {which/that} got me interested in clefts


B: yes, in fact it was Mary {who/?*that} got me interested in clefts


B: yes, in fact it was Mary {who/that} everybody knows got me interested in clefts

• Kayne’s observation may be relatable to Belletti’s recent suggestion that the specifier position of the

relative clause in highest-subject it-clefts is an A–position — which suggests a link with pseudo-

relatives (cf. (76)), as Belletti notes (but she does not develop this in any detail)

(76) j’ai vu Marie qui mangeait une pomme (French)

I’have seen M arie who ate an apple

‘I saw Marie eat an apple’


6 while the claim that the specifier position of the relative clause of highest-subject it-clefts is an

A–position is not obviously correct as a general claim about ALL it-clefts, it is highly plausible that

it is correct on a more modest scale, confined in scope to CONTINUOUS-TO PIC it-clefts

6 in particular, imagine that in the structure of a highest-subject continuous-topic it-cleft, the VALUE

is actually base-generated in SpecCP, with the wh-constituent occupying the SpecIP position and

being bound by the VALUE in SpecCP, as in (77) — a structure that closely resembles Guasti’s (1993)

analysis of pseudorelatives of the type in (76) (which base-generates the ‘head’ of the pseudorelative

in the specifier of ‘AgrCP’, the projection in the C–domain that has an A–specifier position)

(77) [SC [CP VALUE[+human] [CN (*that) [IP who [IN I ...]]]] [Predicate it]]

6 if the VALUE is in SpecCP in highest-subject continuous-topic it-clefts, as in (77), and if SpecCP in

these it-clefts is an A–position (as Guasti 1993 proposed for the SpecCP of pseudorelatives), the ban

on that-relatives with [+human] VALUEs follows immediately on the assumption that a [+human]

occupant of SpecCP qua A–position must agree with C but clashes with a lexicalisation of C as that

(which is incompatible with [+human] — it goes back to the distal demonstrative pronoun, which

cannot refer to humans: he/she/*that talks too much; see Kayne 2008 for a similar line of thought)

• going much further than Belletti in empirical detail, Lambrecht (2002) explicitly establishes a parallel

between pseudorelatives and a particular subtype of clefts in French — esp. constructions like (78)

(so-called avoir-clefts)

(78) a. j’ai une dent qui manque (French)

I’have a tooth that misses

‘I have a tooth missing’

b. y a la téléphone qui sonne

there has the phone that rings

‘the phone is ringing’

6 Lambrecht claims that in both pseudorelatives and clefts of this type the constituent preceding the

embedded clause is the focus of the matrix clause but the topic of the embedded clause

6 Lambrecht uses the label ‘presentational relative construction’ (PRC) for all his examples, including

his avoir clefts and perception constructions with pseudorelatives

6 he observes that ‘the discourse function of this construction type is to introduce a new entity into a

given discourse world and at the same time to express some piece of new information involving this

entity’ (p. 171) — so, crucially, they have in common with continuous-topic it-clefts (and also with

discontinuous/all-new it-clefts) the fact that the relative clause is not presuppositional but new

6 Lambrecht argues that ‘the semantic function of the relative is neither restrictive nor appositive but

PREDICATIVE’ (p. 172; original small caps), an insight that the structure in (77), below, captures rather

handsomely

6 for Lambrecht, in pseudorelatives and avoir-clefts ‘[t]he RC denotatum ... has a FOCUS relation to the

global proposition’ (p. 178) — because, I would argue, it is the relative clause as a whole that serves

as the subject of the specificational copular sentence, as depicted in (77)

6 precisely because it represents the entire CP including the VALUE and the rest of the relative clause

as the subject of the small clause, (77) predicts that the entire CP is the focus of the cleft, not just the

VALUE — after all, it is the subject of the raised predicate of an inverse specificational copular

sentence that qualifies as the focus

6 this is exactly right result for continuous-topic it-clefts, which have the peculiar property that it is not

the VALUE per se that is the focus of the construction but rather the predication established between

the VALUE (qua ‘head’ of the relative clause) and the predicate denoted by the relative clause


27 Chomsky’s example features meet as the verb of the parasitic adjunct clause; but for many English speakers, meet is a once-only predicate (‘making one’s first acquaintance’), so meet fits rather poorly with whenever. I have replaced meet with see, therefore.

28 Thanks especially to Peter Sells and Christina Tortora for their help with these sentences. I should note for the record thatPeter Svenonius has indicated to me that for him (79d) is no worse than (79b,c); so the pattern indicated in the text is not universal.

6 thus (77) allows us to explain both the ban on that in highest-subject continuous-topic it-clefts with

a [+human] VALUE and the information-structural profile of these constructions (as distinct from con-

trastive/stressed-focus it-clefts)

• evidence that who in highest-subject discontinuous-topic it-clefts is in an A–position may come from

parasitic gaps (though I should forewarn the audience that the facts here are tenuous)

6 Chomsky (1986) gives (79a) as grammatical (attributing the observation to Longobardi)27 — and

though (79a) certainly does not meet with universal approval, there are indeed speakers who accept

it marginally

6 (79a) is unambiguously a predicational copular sentence; but speakers who accept (79a) also accept

(to the same marginal degree) the specificational copular sentences in (79b) and (79c)

6 however, even speakers who (marginally) accept (79a–c) tend to reject the it-cleft in (79d)28

(79) a. ?John is a man who [whenever I see pg] looks old

b. ?John is the man who [whenever I see pg] looks old

c. ?the only man who [whenever I see pg] looks old is John

d. *it’s John who [whenever I see pg] looks old

6 the fact that, for the speakers in question, (79d) is significantly worse than the other sentences in (79)

seems to indicate that the relative clause in this it-cleft behaves differently from the relative clause

in other contexts with respect to the ability to license parasitic gaps: only in (79d) does this fail

6 there certainly is no general ban on parasitic gaps in it-clefts: (80) is grammatical, so object clefts

license pgs — the problem with (79d) is specifically the fact that who is the highest subject of the

relative clause in the cleft

(80) it’s John who Mary kissed before scolding pg

6 so if indeed (79d) is to be deemed ungrammatical (for a subset of speakers), then this may be (very

tentatively) chalked up as evidence that the who of highest-subject it-clefts does NOT engage in an

AN–dependency with the subject gap — a conclusion that would be compatible with (77)

[note, however, that in (79d) I did not control for the contrastive vs continuous-topic readings of it-

clefts — in fact, since I did not give my informants a context that explicitly identified (79d) as a

continuous-topic it-cleft, it may be that for the speakers in question, highest-subject it-clefts generally

ban parasitic-gap licensing — a result that would raise further questions that I cannot address here]

• assuming that (77) is on the right track, we should realise immediately, of course, that the story just

told about highest-subject continuous-topic it-clefts with [+human] VALUEs is incomplete

(a) it does not say anything as yet about the fact that that IS allowed in highest-subject

continuous-topic it-clefts with a [–human] VALUE (recall (17B))

(b) nor does it have anything to say about continuous-topic it-clefts whose VALUE is not the

highest subject of the relative clause — and, concomitantly, it does not say anything so far

about Prince’s 1978 observation that that/wh-omission is impossible in continuous-topic it-

clefts (recall (16))


29 See Kayne (2008) on why those does not occur when the VALUE is a plural [–human] noun phrase (cf. *it is Mary’s booksthose got me interested in clefts). While I do not consider Kayne’s remarks conclusive, I have nothing to add to them at this time.I am not committed to the view that that in English relatives is generally a demonstrative: the text suggestion is limited to (17B). Asa matter of fact, I cannot think of any positive evidence to confirm directly that that in (17B) is a demonstrative occupying SpecIP.The fact that that in (17B) resists a full vowel /æ/ in the same way that other instances of that in relative clauses do does not seemto bode well for Kayne’s demonstrative hypothesis. Note, incidentally, that in highest-subject it-clefts with a [–human] VALUE andwith that in the left periphery of the relative clause, parasitic gaps seem to be just as difficult to license as they are in (79d) (*it’sJohn’s car which/that whenever I see pg looks like it’s about to fall apart). There is no evidence, therefore, for AN–movement toSpecCP in (17B), any more than there is in the case of (18B). This is compatible with both text approaches to (17B) with that.

re: (a) recall that that is impossible in (18B) because (i) the VALUE occupies SpecCP, (ii) SpecCP is an A–

position in continuous-topic it-clefts (just as in pseudorelatives), (iii) the VALUE is [+human], but (iv)

that does not tolerate a [+human] constituent in its A–specifier: A–specifiers must match in phi-

features with the heads that they are specifiers of, but that, which plainly is not [+human] (recall the

discussion right below (77), above), cannot match the [+human] feature of its specifier in (18B)/(77)

6 by contrast, a [–human] VALUE in SpecCP qua A–position does not clash in phi-features with that

in C — and as a consequence, there should in principle be nothing wrong with that lexicalising C in

(77) when the VALUE is [–human]

6 with the [–human] VALUE in SpecCP, that lexicalising C, and SpecIP being non-overt (presumably

occupied by pro, as in Guasti’s 1993 analysis of pseudorelatives), we can thus obtain (17B) with that

6 besides presenting a ‘doubly-filled Comp’ effect, however, this derivation raises the non-trivial quest-

ions of how pro is licensed and why that and which cannot co-occur (17B)

6 as a possible alternative, we could follow Kayne (2008) and assume that that in (17B) does not

lexicalise C but is a demonstrative pronoun lexicalising the subject in SpecIP 29

re: (b) the structure in (77) caters specifically for highest-subject continuous-topic it-clefts

6 this is basically a good result: according to Prince (1978:899), continuous-topic it-clefts are vastly

more common (in corpora) with highest subjects than with objects or other constituents

6 Prince was successful, though, in constructing (16b) (repeated on p. 24, above) on the basis of an

attested example with a highest-subject focus; and Kayne (2008) gives a made-up example of a

continuous-topic it-cleft with an object gap in the relative clause: do you know Mary? — yes, in fact

it was Mary {who/?that} I learned linguistics from in the first place

6 so continuous-topic it-clefts are not strictly confined to highest subjects, though they tend to be par-

ticularly common in corpora with highest subjects

6 let us tentatively take this to mean that (77) (which fits the information-structural profile of these

clefts perfectly) is actually correct as the schema for continuous-topic it-clefts in general, and that

non-(highest-)subject continuous-topic it-clefts are ‘squeezed into’ this general schema

6 specifically, let us tentatively assume that non-(highest-)subject continuous-topic it-clefts ARE in fact

(surface appearances notwithstanding) instances of highest-subject continuous-topic it-clefts — but

that the portion of the relative clause in which the VALUE identifies the highest subject is elliptical

(81) a. [SC [CP VALUE[+human] [CN (*that) [IP who is the person [who/that ...]]]]] [Predicate it]]

b. [SC [CP VALUE[+human] [CN (*that) [IP who is the person [who/that ...]]]]] [Predicate it]]

6 (81a,b) present the two logically possible ellipsis choices

6 of these, (81a) makes exactly the same predictions regarding the physical left periphery of the relative

clause as does (77): only who should be allowed with a [+human] VALUE

6 (81b), by contrast, predicts that the physical left periphery of the relative clause should be subject to

exactly the same restrictions that a restrictive relative clause associated with a [+human] predicate

nominal is normally subject to — i.e., (81b) allows that for the same reason that the person that...

is grammatical


6 so via (81b), we can accommodate the possible occurrence of that in non-(highest-)subject continu-

ous-topic it-clefts

6 AND we also get a straightforward handle on the fact that a null left periphery is disallowed in such

clefts (recall Prince’s (16b)):

– (81a) predicts that who MUST show up for the same reason that who must show up in (77)

(basically, the requirement that SpecIP be filled)

– (81b) predicts that either who or that MUST show up because the fact that the matrix is ellip-

tical prevents licensing of a null left periphery in the embedded relative clause (basically for

the same reason that complementiser deletion in the clausal complement of a gapped bridge

verb is impossible: John said (that) Mary had done it, and Bill said *(that) Sue had)

6 Prince’s (1978) observation that continuous-topic it-clefts do not allow a null left periphery is thus

accounted for

• what remains to be said is that, apparently, an elliptical derivation for a highest-subject continuous-

topic it-cleft is unavailable — plainly, if we allowed (81b) as the underlier for highest-subject con-

tinuous-topic it-clefts, we would lose our account of Kayne’s observation in (18B)

6 an elliptical derivation for continuous-topic it-clefts apparently is only available as a last resort —

so only for non-(highest-)subject cases, for which (77) as it stands is not applicable; since (77) itself

takes care directly of highest-subject continuous-topic it-clefts, the more complex source in (81b) is

not available

6 though this makes intuitive sense, formalising it as part of a general theory of the distribution of

elliptical underliers is by no means trivial; I am currently unable to do more than this, however

ReferencesBennis, H. 1986. Gaps and dummies. Dordrecht: Foris.Chomsky, N. 1986. Barriers. Cambridge, MA: MIT Press.Declerck, R. 1988. Studies on copular sentences, clefts and pseudo-clefts. Leuven: Leuven University Press/Dordrecht: Foris.Dikken, M. den. 2005. Specificational copular sentences and pseudoclefts. In H. van Riemsdijk & M. Everaert (eds), The Blackwell

companion to syntax, Vol. IV. Oxford: Blackwell.Dikken, M. den. 2006. Relators and linkers: The syntax of predication, predicate inversion, and copulas. Cambridge, MA: MIT Press.Dikken, M., A. Meinunger & C. Wilder. 2000. Pseudoclefts and ellipsis. Studia Linguistica 54: 41–89.Doetjes, J., G. Rebuschi & A. Rialland. 2004. Cleft sentences. In F. Corblin & H. de Swart (eds), Handbook of French semantics. Stanford:

CSLI Publications. 53–69.É. Kiss, K. 1998. Identificational focus versus information focus. Language 74. 245–73.Erades, P. 1962. Points of modern English syntax. English Studies 43. 136–41.Gryllia, S. & M. Lekakou 2007. Clefts in Cypriot Greek. Ms., Leiden University & Meertens Institute, Amsterdam.Guasti. M.T. 1993. Causative and perception verbs: A comparative study. Turin: Rosenberg & Sellier.Gundel, J. 1977. Where do cleft sentences come from? Language 53. 543–59.Hedberg, N. 2000. The referential status of clefts. Language 76. 891–920.Huber, S. 2002. Es-Clefts und det-Clefts: Zur Syntax, Semantik und Informationsstruktuur von Spaltsätzen im Deutschen und Schwedischen.

Stockholm: Almqvist & Wiksell International.Kaisse, E. 1979. Auxiliary reduction and the derivation of pseudoclefts. Linguistic Inquiry 10. 706–10.Kayne, R. 1981. ECP extensions. Linguistic Inquiry 12. 93–133.Kayne, R. 2008. Why isn’t this a complementizer? Ms., New York University.Koster, J. 2000. Extraposition as parallel construal. Ms., University of Groningen.Kuroda, S.-Y. 1972. The categorical and the thetic judgement. Evidence from Japanese syntax. Foundations of Language 9. 153–85.Lambrecht, K. 2002. Topic, focus and secondary predication: The French presentational relative construction. In C. Beyssade, R. Bok-Ben-

nema, F. Drijkoningen & P. Monachesi (eds), Romance Languages and Linguistic Theory 2000. Amsterdam: Benjamins. 171–212.Moro, A. 1997. The raising of predicates. Cambridge: Cambridge University Press.Percus, O. 1997. Prying open the cleft. Proceedings of NELS 27. Amherst: GLSA. 337–51.Prince, E. 1978. A comparison of wh-clefts and it-clefts in discourse. Language 54. 883–906.Reeve, M. 2007. Relatives and pronouns in the English cleft construction. Ms., University College London.Rijkhoek, P. 1998. On degree phrases and result clauses. Ph.D. dissertation, University of Groningen.Sasse, H.-J. 1987. The thetic/categorical distinction revisited. Linguistics 25. 511–80.Vries, Mark de. 1999. Extraposition of relative clauses as specifying coordination. In Proceedings of ConSOLE 7.Williams, E. 1983. Semantic vs. syntactic categories. Linguistics and Philosophy 6. 423–46.

A Cleft Palette - Graduate Center, CUNY · A Cleft Palette On the lands cape of cleft co nstr uctio...

Documents

Transcript of A Cleft Palette - Graduate Center, CUNY · A Cleft Palette On the lands cape of cleft co nstr uctio...