Review of Relators and Linkers
|AUTHOR: Marcel den Dikken, City University of New York
TITLE: Relators and Linkers
SUBTITLE: A Study of Predication, Predicate Inversion, and Copulas
SERIES Title: Linguistic Inquiry Monographs
PUBLISHER: MIT Press
Pavel Grashchenkov, unaffiliated researcher, Moscow, Russia.
Chapter 1: Introduction.
Chapter 1 presents in brief the main proposal of the book. According to den
Dikken, every predication represents a nondirectional structure with a
special ''pivot'', instantiated by RELATOR. RELATOR surfaces as a copular in
the nominal predication but may be also found in the domain of noun phrases:
(1) een beer van een vent
a bear of a bloke
den Dikken defines RELATORs (English 'as', Dutch 'als' etc.) as elements
that ''mediate the syntactic relationships between the predicate and its
subject''. Then, in examples like (2) 'to be' (which is not omissible due
to locality and equidistance) is a LINKER, used to make the predicate
(2) Imogen considers the best candidate *(to be) Brian.
Chapter 2: The Syntactic Configuration of Predication.
Whereas the predicate is defined as a ''syntactic constituent that expresses
a property ascribed to the subject'', den Dikken's understanding of subject
differs from the traditional one: under this term the traditional
''grammatical subjects'' (the surface subjects of transitive, unergative and
unaccusative verbs) are unified with the Topic in Topic-Comment structure,
for instance, 'flower' in 'a beautiful flower'. All predication
relationships arise from a structure in which the predicate and the subject
are combined by means of a RELATOR. Then, ''The locality of predication''
(3) The RELATOR accommodates the predicate and the subject in its MINIMAL
One of the main claims of the author is that predication relationships are
asymmetrical and non-directional and are usually realized as a
Predicate-complement structure, (4.a) or Predicate-specifier structure, (4.b):
(4.a) [RP [XP SUBJECT] [R' RELATOR [YP PREDICATE]]
(4.b) [RP [XP PREDICATE] [R' RELATOR [YP SUBJECT]]
The role of RELATOR in den Dikken's system is similar to that of Pred of
Bowers (1993). As opposed to Pred, which is never instantiated by any
lexical or functional category, RELATOR is a placeholder for any head in
syntactic structure (T, D, 'be', 'as', 'for'...). RELATOR also
participates in the formation of S(mall) C(lauses): instead of VP-shell
structure, proposed for sentences like
(5) Imogen put the book on the shelf
In Hale and Kayser (1993), den Dikken proposes that the lexical verb
('put') occupies the higher V position whereas the RELATOR heads the SC:
(6) [VP put [RP [ the book] [R' RELATOR [on the shelf]]].
A RELATOR is a functional head, which does not assign a theta-role. It can
introduce Top and Foc noun phrases and mediates the modifier and modifee in
case of adjectives and adverbs. RELATOR may be also found among examples
that are not easy to capture in more traditional terms:
(7) This butterfly is [RP [AP big] [R' for [DP a butterfly]].
The RELATOR plays also a significant role in active / passive alternation
(as well as in causatives, perfect and other complex event structures).
Whereas active verb phrases are derived with the ''standard'' subject in the
specifier position and RELATOR=v, in the passive the Spec position is
occupied with the VP and RELATOR='by':
(8.a) [vP [DP Imogen] [v' [v ( (RELATOR)] [VP kissed Brian]]]
(8.b) [RP [VP kissed Brian] [R' [P by (RELATOR)] [DP Imogen]]]
Chapter 3: Small Clauses and Copular Sentences.
Among the most interesting approaches to the SC structure, den Dikken
distinguishes the following three: flat structure, Williams (1980); complex
predicate, Johnson (1991) and small-clause analysis, Kayne (1984).
The author supposes that SC constitutes the single set of phenomena with
the causatives ('Je laisse Imogen embrasser Brian'), ditransitives ('Brian
hung the shirt on the line') and some other constructions, all them making
use of the RELATOR. According to den Dikken ''all small clauses are
projections of a functional head'', thus he argues against the ''bare'' SC
In particular, for equative sentences den Dikken proposes a structure in
which one of the two equative noun phrases is derived as a free relative,
having undergone Predicate Inversion:
(9) [TP [PRED PRO-PREDICATE] [CP Op(i) [C ( [RP Cicero [RELATOR1 ( t(i)]]]]]
[T+RELATOR2='be' [RP Tully [t(k) t(j)]]]
'Tully is what Cicero is' ('What Cicero is, is Tully')
Another important generalization concerning copular sentences is that ''all
copular sentences feature a small clause in their underlying syntactic
Chapter 4, Predicate Inversion: Why and How
Comparing inverted and uninverted predicates, the author observes that
''(i)n uninverted predications a heavy stress may fall on the last element
and propagate up to higher nodes in the syntactic tree '', (10.a), that is
not the case in inverted predications, (10.b):
(10.a) She does not consider him to be the [DEVIL INCARNATE].
(10.b) *She does not consider the best candidate to [have been LYING].
Then, it is observed that A'-extraction of the postverbal subject of
Copular Inversion constructions is impossible:
(11.a) I think the best candidate is this man.
(11.b) *Which man do you think the best candidate is?
In the same time, in the constructions that were previously argued to be
derived by means of RELATOR and predicate-specifier structure, such
extraction is possible:
(12) ?[A five-year old], Brian would actually be clever for t.
It follows then that Copular Inversion constructions are not derived as
predicate-specifier structures as one could suppose:
(13) be [RP [DP the best candidate] [RELATOR [DP Brian]]]
The additional argument that forces us to reject (13) is that whereas the
postcopular subject in Copular Inversion constructions cannot project focus
up the tree, see (10.b), it can do this in predicate specifier constructions:
(14) Imogen said Brian is clever for a FIVE-YEAR-OLD.
The author then argues that Predicate Inversion is an instance of A-movement:
(15.a) It was claimed that the best candidate was Brian.
(15.b) The best candidate was claimed to be Brian.
According to den Dikken, equative nominal predications ('Cicero is Tully')
as well as specificational predications ('Brian is an excellent doctor')
are always instances of Predicate Inversion including reduced free
relatives (see (9)). The rationale for Predicate Inversion is stated as
(16) Predicate Inversion involves A-movement to subject position triggered
by the need to license an empty predicate head.
The same analysis applies to Locative Inversion. But how does the existence
of inverted and uninverted variants satisfy the Economy Condition? den
Dikken argues that inverted sentences differ in that they feature an
additional means, i.e. a pro-predicate, hence the numeration in the two
derivations does not coincide and they can not be compared on Economy grounds.
Another important claim made in this chapter is that SC is a phase. Hence,
the P(hase) (I)mpenetrability (C)ondition follows. But if PIC is valid for
SC, how do Predicate and Locative Inversion become possible, since inverted
structures (see 17-18.b) explore the raising of complements (=predicates)
to the subject position?
(17.a) [T [RP Brian [RELATOR [the best candidate]]]]
(17.b) [T [RP Brian [RELATOR [NULL PRO-PREDICATE [CP Op(i) [C _ [RP t(i)
[RELATOR [the best candidate]]]]]]]]]
(18.a) [T [RP a picture of Imogen [RELATOR [on the wall]]]]
(18.b) [T [RP a picture of Imogen [RELATOR [NULL PRO-PREDICATE]]]]
Moreover, we are faced with another problem: from the point of view of
Relativized Minimality, subjects in inverted constructions seem better
candidates for movement to Spec, TP than predicates. To solve both these
problems, den Dikken proposes that RELATOR in SC may raise to some external
functional head, which he calls LINKER. This has two aims: first,
adjunction of the lower head (RELATOR) to the higher one (LINKER) deletes
the (borders of) the lower phase; second, it makes the predicate and the
subject of SC equidistant to the landing sites outside this SC.
Another important question is why A'-extraction of postverbal subject is
regularly ruled out:
(19.a) Imogen thinks that the best candidate is Brian.
(19.b) *Which guy Imogen thinks that the best candidate is?
(20.a) Imogen thinks that on this wall hung a picture of Brian.
(20.b) *Whose picture does Imogen think that on this wall hung?
To explain this, den Dikken notes that:
(21.a) The subject of Predicate Inversion constructions is in a FOCUS position.
(21.b) A'-extraction of a constituent occupying a FOCUS position is impossible.
As for the LINKERs, not only copulas can play this role, but null aspectual
heads as well:
(22) If Bill has an alibi for 6 p.m., that makes [AspP the murderer
Chapter 5, Predication and Predicate Inversion inside the Nominal Phrase
Chapter 5 proposes a detailed investigation of predication and Predicate
Inversion in the nominal domain. One of the main topics of the chapter is
the structure of the Qualitative Binominal Noun Phrase (QBNP): 'a jewel of
a village', 'an idiot of a doctor'. The first example is attributed to the
''comparative'' QBNP, the second – to the ''attributive'' QBNP. These two types
of constructions have differences not only in their semantics but also in
syntactic realization: whereas the attributive QBNP allows for an
juxtapositional correlate, the comparative one does not (in (23.a) only the
reading 'a village with jewels' is available):
(23.a) (*) a jewel village
(23.b) an idiot doctor
den Dikken argues that attributive QBNPs have predicate-specifier
structure, (partially) patterning with constructions such as 'an idiot of a
(24.a) [DP [RP [NumP an idiot] [RELATOR=of [NumP a doctor]]]]
(24.b) [NumP an [RP [NP idiot] [RELATOR=of [NP doctor]]]]
(24.c) [RP [NumP an idiot] [RELATOR=as [NumP a doctor]]]
Two significant observations must be made here: first, the RP structure
(24.a) repeats that of (24.c) with the only difference being that in the
former case it is embedded in the DP shell whereas in the latter it is not.
As a consequence, the 'of' realization of the (nominal) RELATOR is observed
in (24.a) but not in (24.c). Second, there is evidence that the subject and
the predicate in these structures are NumPs headed by the indefinite
articles as opposed to that of (24.b).
Comparative QBNPs originate as predicate-complement structures, in which
'of' serves as a LINKER above the SC with the subject and predicate. The
(dummy) RELATOR then raises to the LINKER that surface as 'of'. The SC
predicate raises to the Spec of the LINKER phrase, both predicate and
subject noun phrases having status of NumP:
(25) [FP [NumP PREDICATE](j) F(=LINKER)+RELATOR(i) [RP [NumP SUBJECT] [R'
It is stated that, as in other instances of the Predicate Inversion,
raising of the predicate happens only when the RELATOR is an empty head
(abstract SIMILAR head in the above example); when it is phonetically
filled, the raising does not happen:
(26) a village like a jewel
'Of' is treated as a nominal copular, analogous to 'be'. QBNPs that clearly
depart from the possessive noun phrases (for instance, the latter but not
the former allow for extraction, 'Of a girl he saw a picture yesterday' may
have only possessive reading) is a good example of this claim. According to
(27.a) The RELATOR is realized as 'of' in the complement of D.
(27.b) The RELATOR is realized as 'be' in the complement of T or Asp.
(27.c) The RELATOR is realized as 'as', 'for', 'like' in the complement of
the RELATOR or specific Vs.
(28.a) The LINKER is realized as 'of' in DP.
(28.b) The LINKER is realized as 'be' in TP.
(28.c) The RELATOR is null in Asp.
Finally, it is argued that possessive noun phrases constitute subject –
predicate relationships inside the RP with the dative / 'of' RELATORs as
well. If we assume this, we can explain (29.a) as an instance of
Predicative Inversion of (29.b):
(29.a) Brian is [a man of means].
(29.b) the [means of that man]
Chapter 6, Predication, RELATORs and LINKERs
This chapter summarizes the main results of the book.
As mentioned in the beginning of the book, some ideas arise from the works
of Teun Hoekstra. Thus, the treatment of 'have' in the perfect tense as a
result of the conflation of 'be' with the dummy preposition and the
analysis of nominal possessives as a predication with (dummy dative)
prepositions (''Benveniste hypothesis'') follow the analysis developed by
Hoekstra (see, for instance, Hoekstra, 2004). den Dikken's notion of
RELATOR and its function in predication are also close to Bower's ''Pred''.
At the same time, some aspects of den Dikken's theory resemble the approach
to the syntax of predication developed by Pesetsky and Torrego. For
example, according to Pesetsky & Torrego (2004), P is an instance of
T(ense) and this is what is reflected in the idea of RELATOR. Thus, the
book discusses problems that were in the focus of attention of topmost
linguists in the last two decades and presents a new approach to these
problems. This, however, does not preserve the theory proposed by den
Dikken from some weak points. I will first discuss some general
observations, and then make some more specific ones.
1) The idea of RELATOR is not very clear. If the notion RELATOR adds some
syntactic or semantic properties to the heads that are grouped under this
term (T, v, D, Top, Adj) it is not clear what these properties really are.
If it is a mere taxonomic unification of a group of functional heads under
the general term, it does not seem useful.
2) Having the RELATOR head the predication in which the substantial part
(complement of the RELATOR) is filled with another RELATOR (phrase) is not
clear from the point of view of semantics (see, for instance, examples (49)
at p.39 and (13) at p.167).
3) It is also argued by the author that ''standard'' predication is parallel
to adjectival / adverbial modification: both structures arise from RELATOR
phrases. Then, the adjunction as an independent operation (opposed to
merging of arguments) must be eliminated from syntax.
4) By the same reasoning the opposition between argument and topic/focus
must be abolished. But adjunct and topic/focus differ crucially from
arguments and predicates.
5) The motivation for LINKERs (proposed on p. 116) seems also a bit
unnatural: LINKERs are introduced to satisfy the minimalist requirements
(namely, the Relativized Minimality). But if it is the only reason that
calls for such a construct, this does not conform to the spirit (although
does conform to the letter) of Minimalism.
1) On p.17 the author mentions that coordinate phrases may be analyzed as
RPs, where 'and' serves as a RELATOR. This analysis should resolve the
long-standing problem of the internal structure of conjoined constituents.
But it soon poses problems, since one of the conjuncts is assigned the
status of subject with respect to the predicate when this status is already
assigned to another conjunct.
2) The author notes that such a structure gives ''a straightforward
reflection of the semantics of coordination, with the coordinator
represented as a connective that takes one set-denoting linguistic
expression, delivering the intersection of the two sets''. It may be true
with regard to adjectival coordination (''nice and easy''), but this could
not be applied to conjoined structures with two noun phrases or verbs: both
'boys and girls' and 'saw and won' cannot be viewed as intersections.
3) On p.23: ''For transitive sentences like 'Imogen kissed Brian' we have
assumed (see section 2.2.4) a semantic representation in terms of the
intersection of two sets, one denoted by the subject (Imogen) and one by
the predicate (kissed Brian).'' It's a strange observation since referents
of proper names are not sets of individuals but unique objects.
4) In the example (33) at p.33, (and some subsequent examples) one finds
that in the structure, proposed for the sentence 'Imogen is beautiful as a
dancer' one RELATOR phrase occurs in the complement position of the other.
This poses serious questions about the nature of the category RELATOR: if
it is no more than ''mediator'', then (33) must be illicit, if it functions
not only to ''relate'' the predicate and its subject, it is not clear for
what reasons RELATOR should replace more standard I(T), D, Top, Adj and
other specified functional heads.
5) The principle stated on p. 50: ''A predicate-sprecifier structures with
adjectival, nominal, and prepositional predicates receive an attributive
interpretation'' (i.e. constructions where predicates are merged as
specifiers never happen to surface as predicates) seems quite interesting,
but, unfortunately, it lacks (like some other general statements in the
book) necessary explanations.
6) On p.149, applying his LINKER-based approach to Norwegian data, the
author notes that LINKERs are merged only as a last resort. If the
structure allows for Predicative Inversion without assistance of LINKERs,
it should be created without them. In particular, if the phase domain can
be extended via raising of the SC head to RELATOR, there is no need for
LINKER to be merged. But, however, the syntactic tool whose only function
is domain extending does not seem necessary. In my opinion, in introducing
a new tool in the syntactic inventory we should take care for their
interpretive properties as well.
7) Considering the structure of attributive and comparative QBNPs at p.
200, den Dikken proposes that ''head'' noun phrases move to the Spec of
Dem(onstrative) phrase. Such an approach, under which various Specs
projected by functional heads inside DP serve as landing sites, when a
movement must happen, become generally adopted after works by Kayne, Cinque
and other researchers. However, it seems much more consistent to preserve
Spec position for the arguments or predicate / subject relations following
den Dikken's own proposal.
8) On p. 204, discussing the ''spurious'' articles in QBNPs, den Dikken notes
that the ''D(efineteness) A(greement)'' effect holds wherever in QBNPs. The
natural question arises then: why doesn't it hold in ordinary possessive
phrases, as is exemplified with (75), p. 202? If the reason here is the
presence of the a RELATOR or a LINKER (and Predicative Inversion in the
latter case), then, should one expect for Hebrew or Arabic noun phrases, in
which DA holds wherever, that RELATORs and LINKERs participate in every DP?
9) Hungarian data, set out on pp. 217-218, calls for discussion of Russian
(and most of Slavic, as far as I know). Thus, in Hungarian nominal
predication with a 3rd person subject no copular should be used. At the
same time, Hungarian comparative QBNPs also do not use 'of'-like elements:
(30) csoda egy könyv
wonder a book
'wonder of a book'
It is then argued that Hungarian lacks overt LINKERs, but resorts to
Predicate Inversion both in nominal predication and QBNP. Russian also
does not make use of a copular in the present tense nominal predications
and it also has Predicate Inversion, but it totally lacks QBNP-like noun
(31.a) Pet'a uchitel'.
'Peter is a teacher.'
(31.b) V sadu (byli) ludi.
in garden (were) people
'In the garden there are (were) people.'
(32) #idiot doctora
'an idiot of a doctor'
(not QBNP, valid only as ordinary possessive)
It seems for me that Slavic data could not be incorporated in den Dikken's
theory so easily as other material discussed in the book.
10) Finally, the treatment of indefinite noun phrases (for instance, in
structures like 24) as NumPs headed by indefinite articles requires
Despite these points, the book proposes a very interesting analysis of a
large group of linguistic phenomena: Small Clauses, Equative Clauses,
Predicate Inversion, Quantitative Binominal Noun Phrases etc., which
intuitively seem to share some common properties. Den Dikken's efforts to
capture this intuition and to spell it out in formal terms give birth to a
new approach to the syntax of predication and the noun phrase and this
enriches modern syntactic theory. The book will be of great interest for
every syntactician, but especially for those who deal with ''nominal domain''
of syntactic theory.
Bowers, John (1993) The Syntax of Predication., Linguistic Inquiry 24: 591-656.
Hale, Kenneth & Samuel Kayser (1993) ''On Argument Structure and the Lexical
Expression of Syntactic Relations'', in Hale, Kenneth and Samuel Kayser
(eds.) The View from Building 20, MIT Press.
Hoekstra, Teun (2004) ''Arguments and Structure: Studies on the
Architecture of the Sentence'', Sybesma, Rint; Barbiers, Sjef; Doetjes,
Jenny; Den Dikken, Marcel; Postma, Gertjan; Wyngaerd, Guido Vanden (eds.),
Studies in Generative Grammar 67, Mouton de Gruyter, Berlin, New York.
Johnson, Kyle (1991) Object Positions. Natural Language and Linguistic
Theory, 9: 577-636.
Kayne, Richard (1984) Connectedness and Binary Branching, Dordrecht: Foris.
Pesetsky, David & Esther Torrego (2004) ''Tense, Case, and the Nature of
Syntactic Categories'', Guéron & Lecarme (eds.), The Syntax of Time,
Cambridge (Mass.): MIT Press, 495-537.
Williams, Edwin (1980) Predication, Linguistic Inquiry 11: 81-114.
| ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Pavel Grashchenkov graduated from the Department of Theoretical and Applied
Linguistics at Moscow State University. His M.A. thesis "Typology of
Numeral Phrases" deals with the morphosyntax of noun phrases with numerals
in the languages of the world. His Ph.D. thesis "Syntax and Typology of
Genitives", defended in 2006, concerns the syntactic position and status of
genitive noun phrases. His academic interests include the internal syntax
of DP, Predicate Nominals, Case, Part-of-Speech Systems, and Typology
(Caucasian, Turkic, Slavic, Finno-Ugric).