LINGUIST List 16.2534
Thu Sep 01 2005
Review: Syntax/Morphology: Hoekstra (2004)
Editor for this issue: Megan Zdrojkowski
Teun Hoekstra: Arguments and Structure
Message 1: Teun Hoekstra: Arguments and Structure
Pavel Grashchenkov <gra-paul
Teun Hoekstra: Arguments and Structure
AUTHOR: Hoekstra, Teun
EDITORS: Sybesma, Rint; Barbiers, Sjef; Doetjes, Jenny; Den Dikken,
Marcel; Postma, Gertjan; Wyngaerd, Guido Vanden
TITLE: Arguments and Structure
SUBTITLE: Studies on the Architecture of the Sentence
SERIES: Studies in Generative Grammar 67
PUBLISHER: Mouton de Gruyter
Announced at http://linguistlist.org/issues/16/16-992.html
Pavel Grashchenkov, PhD student, Department of Theoretical and Applied
Linguistics, Moscow State University, Moscow, Russia.
The book under review is a collection of papers by one of the leading
European generative linguists of the last several decades, Teun Hoekstra
(1953-1998). The volume aims to acquaint the reader with Hoekstra's views
on a very wide area of grammatical phenomena. The book consists of four
parts: "Argument Structure", "T-chains", "The morphosyntax of verbal and
nominal projections", "Small Clauses" and contains fourteen articles
attributed to the last twenty years of Teun Hoekstra's life and work. Most
of these articles have not been published previously.
Part I. Argument structure
Possession and Transitivity
This article considers syntax and semantics of verbs HAVE and BE.
Following Kayne (1994), Teun Hoekstra supposes that HAVE is derived from
BE via incorporation ("Benveniste hypothesis"). All the instances of BE
(copula, equative, passive, progressive, existential, main verb) may have
the same underlying structure. The main reason to use BE in all these cases
is that inflectional information can not be expressed on the complement XP
According to Hoekstra, Agr(eement) head is an inclusion operator. Agr is
also present in BE-copula sentences, so that the whole complement of BE
constitutes a small clause (SC):
(1) Agr ... (BE / HAVE) [Spec F [SC [DP] Agr [PP] ]], Where F is some
Then both (2.a) and (2.b) may be derived by movement of a subject DP or XP
complement of SC to the main clause subject position due to specificity.
(2) a. John was the cause of all the trouble.
b. The cause of all the trouble was John.
HAVE is treated as a semantic and syntactic inverse of BE. Languages have
either BE or HAVE possessive constructions. In the case of BE constructions,
for instance, in eastern Dutch, we have the locative inversion like
in (2) above. Dative here is an underlying PP adjunct in SC.
Hem zijn de handen vies
him are the hands dirty
In HAVE languages underlying structure is the same as in (1), but PP first
moves to Spec-F position, where the prepositional element incorporates to
BE (resulting in HAVE) and then the "residual DP" (possessor) moves to the
matrix subject position. SC subject (possessum) moves to the Agr-O(bject)
associated with HAVE.
The accusativity pattern, either in auxiliary or in possessive sentences,
is due to the possibility of HAVE to license Agr-O. BE can not license Agr-
O (and accusative) hence both in passives and in possessive sentences with
BE one finds no accusative.
In addition to constructions with BE and HAVE as main verbs, some other
constructions are at issue: copula use of both verbs, possessive DPs,
modal sentences with nominative (English) and dative (Hungarian),
ditransitive verbs. The resulting structure allows to build a unifying
account of BE and HAVE in different types of constructions.
The Indirect Object: Its Status and Place
This paper was written in 1978 and was executed in transformational terms.
It considers contemporary approaches to ditransitive clauses, their word
order and argument marking. The analysis at issue was implemented by quite
technical rules (deletion of prepositions and reordering of remaining
NPs). These approaches, however, faced with some problems. First, as is
argued by Hoekstra, indirect objects are both semantically and
syntactically less close to V (and are generated under V'' and V'
correspondingly, against Jackendoff's proposal). This contradicts the fact
that indirect object NP can bind an anaphor in DO (direct object). Another
problem with binding is that indirect objects PPs may bind DOs:
Ik geef aan de kinderen elkaars tekeningen
I give to the children each others' drawing
'I give to the children each others' drawing.'
One of the main problems for the analysis of IO is its prepositional
realization that must be an obstacle to the binding of DO by IO
antecedent "through" the preposition (and is not in fact).
Examining causative constructions, Hoekstra considers bisentential
analysis for these predications. Here we are faced with another problem:
Dutch (as opposed to French) does not always code subjects of
transitive "internal" predications as IOs (indirect objects), but
sometimes makes use of DOs. The choice here depends on lexical properties
of the verb in subordinate clause. Thus, a problem of the distribution of
(in)direct objects in causatives arises.
To explain all these facts, Hoekstra adopted bisentential analysis in the
spirit of Lexical Decomposition. He suggests that constructions in (5)
include two verbs (and hence two clauses): one of which is the embedded
lexical verb and another is the component CAUSE that may (5.b) or may not
(5.a) surface as an independent lexeme in final structure:
Ik geef (aan) Mary een boek.
'I give Mary a book.'
Ik laat (door) de slager mijn koe slachten.
'I have the butcher slaughter my cow.' /
'I have my cow slaughtered by the butcher.'
"Mary" and "the butcher" are underlying subjects and may surface as NPs or
PPs. As subjects of some embedded predication they may bind the objects "book"
and "cow". They are originated as NPs in the embedded VPs, but may end up
in derivation either as ("dative") NPs or undergo the optional rule of
passivization of the whole internal VP, which results in the overt
preposition ("by" or "to").
Hoekstra also tries to solve the problem of the choice between indirect
and direct objects introducing feature "control": [+control] embedded
verbs show up as DOs, whereas [-control] verbs become IOs (cf. the notion
of "affectedness" in present theories).
The main result of this research may be articulated as follows: "the
relation IO does not exist". It is not an exaggeration to say that some
serious steps were made by Hoekstra towards the present understanding of
indirect object and the inner structure of VP.
Categories and Arguments
The goal of the paper is to propose a universal analysis of clause
structure on the grounds of some general concepts of the lexicon. These
concepts are taken to be merely nominal. Verbs as syntactic items are
always composed of some F-head and lexical material (which can be A or
N). Thus (6.a) is a result of incorporation of the adjective "clear" in F.
(6) a. The screen cleared.
b. John cleared the screen.
In case of transitive clauses, see (6.b), the only difference is that
besides "screen", the SC also contains the abstract preposition P
(relator) (AgrS = Agreement Subject).
(7) AgrS - T - AgrO - F [SC [AP [the screen] clear] P [John] ]
(7) is reorganized to build (6.b) in the following steps: P incorporates
into F, "John" moves to SpecF and then to SpecAgrS, F+P moves to
AgrO, "the screen" moves to SpecAgrO to check its accusative. Hoekstra
here extended his analysis of HAVE from the first article on all the
lexical classes of Vs. He also advocates Stowell's (1981) approach to phrase
structure, under which phrases of every category may have a subject and
argues against Hale and Keyser's (1993) theory of argument structure in
which AP and PP, being predicates, do not take subjects.
The Active-Passive Configuration
In this paper Hoekstra proceeds the investigation of possible consequences
of his Strict Separation Hypothesis and applies the approach articulated
in the previous articles to passive sentences.
(8) Strict Separation Hypothesis: L-primitives are exhaustively
characterized in terms of features corresponding to ontological classes of
individuals. F-primitives have no features that denote ontological classes
What follows from (8) is that verbs in the underlying structure are
present only as some abstract F heads, which supply sentences with
grammatical features of tense, aspect etc. The lexicon contains As, Ns and
Hoekstra relates transitivity (and accusative case licensing) to
possibility of incorporation of some abstract preposition P. Thus we have
an explanation for the Burzio generalization: both projection of the external
argument and licensing of accusative are due to the phenomenon of abstract
P. When P can not be incorporated, the external argument may receive only
oblique marking (via overt P) and no accusative can be assigned, see (7)
The article is attributed to 1986 and was not previously published. The
author poses some questions that have been somehow ignored in previous
research, as for instance: what is the difference between the passive
and perfect participles? To answer this and some other questions, he
proposes his own approach to verbal morphology based to some extent on the
traditional analyses. He states that the external argument is no doubt
present in the thematic grid in accordance with the following principle:
(9) Participial Morphology (PM) bears the external argument role iff it
has case (cf. Roberts 1985).
Considering perfect constructions (10.a) he observes that they have only
agentive interpretation, whereas in constructions like (10.b) "we" does
not obligatorily have agentive interpretation:
(10) a. We have hidden fugitives.
b. We have fugitives hidden.
In the latter case there is a whole small clause, the subject of which
receives the case from HAVE. In the former example HAVE assigns case to PM
and thus there remains a possibility for V ("hide") to assign case to the
internal argument "fugitives". In this case (i.e. in perfect clauses) PM
and subjects constitute an argument chain, hence only the agentive
interpretation is allowed.
The crucial role of HAVE in licensing PM-argument can be traced in the
prenominal use of unergative verbs. In this case PM receives thematic role
from the verb and hence must have a case assigner, which is absent in these
examples. The ungrammaticality of (11) follows from (9):
(11) *a worked housewife / *the worked children / ...
Under this analysis passive and perfect participles represent the same
item, properties of which depend on the syntactic environment: it may be
used to construct the perfect clauses with the auxiliary HAVE licensing PM
or passive clauses with BE, which has no such possibility.
Hoekstra observes that a similar analysis may be applied to infinitives,
where PRO(noun) can be treated as analogous to the overtly absent external
argument in passives. He argues that, just like in the case of PM, it is
I(infinitival) M(orphology) that bears an argument role and requires case
(thus we don't need PRO any more).
To support his analysis of verbal morphology, Hoekstra examines examples
of Dutch impersonal passives, Icelandic quirky case,
French "etre" / "avoir" distribution, Norwegian infinitives and some
Part II. T-chains
Why Kaatje Was Not Heard Sing a Song (with Hans Bennis)
The starting point for the paper is the ungrammaticality of passivization
in clauses headed by perception verbs, which also take sentential
(12) *Kaatje was heard sing a song.
The ungrammaticality of (12) is even more surprising since both ECM (13.a)
and SC (13.b) constructions do allow passivization:
(13) a. Kaatje was believed to have sung a song.
b. Kaatje was considered a good singer.
Hoekstra and Bennis propose an explanation based on the theory of T-
chains. According to Pollock (1989), the verbs may surface in a language
in one of the following positions:
(14) [ C [ T [ Agr [ V ]]]]
T-chains are subject to the following constraints:
(15) T-linking: A verb must be identified by tense.
(16) T-chains may vary across languages on two parameters:
a. the base position of tense;
b. the way in which the chain is established: by verb movement or by
Hoekstra and Bennis argue that in (17-18) and both (a) and (b) SCs are
tenseless AgrPs. Passivization is absolutely grammatical when Agr has an
AP or PP complements, but impossible with VP complements:
(17) a. John was considered [t foolish].
b. He was kicked [t off the street].
c. *John was heard [t sing a song].
Hoektrsa argues that unlike English, where T-chaining between matrix and
embedded clauses is established by percolation, Dutch forms T-chains by
movement: V-raising (18.a) or (non-)finite complement extraposition (18.b).
dat Jan [een appel t] moet eten
that John an apple must eat
dat Jan belooft [(om) een appel eten]
that John believe for an apple eat
The ungrammaticality of (12) in both Dutch and English is explained by the
non-tense (and, moreover nominal) nature of passive participles, which
serves as a "barrier" and does not allow movement of the embedded verb
through the matrix V in Dutch and the similar T-percolation in English.
T-Chains and Auxiliaries (with Jacqueline Guéron)
The paper investigates the syntax of verbs which can take CP, IP or VP
complements. The authors define auxiliaries as a class of verbs that can
take VP complements whereas full verbs may govern CP, IP, NP or zero
complements. What straightforwardly follows from this statement is that
auxiliaries do not assign theta-roles. Then, VPs differ from the other
types of complements in that they normally do not contain subjects and
hence allow movement of internal NPs to the domain of the matrix verb.
Following Kayne (1985), Hoekstra and Gueron suggest that even in clauses
where only auxiliaries are present there may be more than one embedded VP.
Allowing the possibility of VP1,2 and movement through the VP2 adjoined
position is the only way to avoid the ECP (Empty Category Principle)
violation. Adjunction to VP is thus the SC subject position resulting in
(19) NP(i) V1 [VP2 e(i) [VP2 V2 e(i) ]]
According to Hoekstra and Gueron, "a temporal marker in INFL(ectional)
attributes a Temporal index (T-index) to a VP it governs". This creates T-
(20) T(k) - AUX(k) - [VP(k) V(K)], where k represents the T-index.
Every CP has its own T-index that explains the ungrammaticality of
examples in which the imbedded verb selects the auxiliary of the matrix
(21) Jean a / *est su que Pier est venu.
VPs must receive T(emporal)-roles just like NPs receive theta-roles. An XP
which is assigned T / theta role is a Complete Thematic Constituent (CTC):
all theta roles associated to X are assigned internal to XP. Then, the
verb "etre", which does assign T-role, governs a CTC, whereas "avoir" does
not assign T-role and admit percolation of the agent role of the embedded
V up to the matrix subject.
Causative constructions, resulting in clitic climbing in French (22.a) and
presence of past participle in Italian (22.b) allow one to conclude that
causatives also take VP complements. They assign T-roles to VP complement
and accusative case to their subjects.
(22) a. Je le(i) fais [VP e(i) voir e(i)]
b. I libri(i) furonto [VP e(i) fatti [VP e(i) leggere e(i) ]]
The same analysis may be applied to modal verbs, complements of which are
also VPs. This holds in Italian but not in French, where modal verbs do
not allow for clitic climbing and are always incompatible with "être". The
embedded phrases in French modal sentences hence must be regarded as CPs.
Clitics in Romance and the Study of Head Movement
Here Hoekstra focuses on differences between main and auxiliary verbs.
According to his proposal, main verbs introduce an e(vent)-role, whereas
auxiliaries do not have e-role (they do not denote, but modify an event).
It is further argued that Tense must be locally linked to a unique e.
Clitic placement obeys the following constraint:
(23) Attach a clitic to the Tense that licenses its governor.
Italian "volere" differs from its French counterpart in that the former
but not the latter can be either main or auxiliary. This difference is due
to verb raising to Agr (higher Tense), which is obligatory in Italian and
is not possible in French in the case of verbs without overt agreement. Some
XP barrier between Agr and T does not allow this in French.
Thus, the configuration in which Tense from the main clause would be
linked to the embedded verb is also barred by the XP barrier. Hence the only
possible use of "vouloir" is that, which is linked to Tense of the matrix
clause, i.e. the use of a main verb. We then can explain the absence of
clitic climbing to "vouloir" in French: in sentences with "vouloir" there
are always two (matrix and embedded) Tenses, hence no climbing happens
(and the opposite holds for Italian).
As Hoekstra concludes, clitic climbing in Romance is not a successive
cyclic, but a long distance movement.
ECP, Tense and Islands
The point of departure here is the following examples, where (24.a)
demonstrates the well-known that-trace phenomenon:
(24) a *Who do you think that bought this novel?
b I don't think that anyone will be arrested.
As can be seen from (24), that-trace works on S-structure but becomes
irrelevant in case of L(ogical) F(orm)-movement. However, for instance, in
French and Dutch that-trace is not found. It is puzzling then, that in
these languages extraction from negative and factive islands is prohibited
whereas in English extraction from negatives is acceptable:
(25) ?Who don't you believe would buy this book?
Factive islands display the same properties as negative in the case of direct
object and adjunct extraction, but extraction of the subject from factive
islands is not possible:
(26) *Who do you regret (that) bought this book.
For languages, where that-trace effects are not valid, it is argued that the
trace is licensed by the movement of V to C position at LF. Such V2
movement allows for the trace to be head governed by V. Moreover, EV2 and
embedded V2 are overt in some Germanic languages.
To explain the ungrammaticality of extraction from sentences with "regret"
and some other factive verbs, Hoekstra suggests that factive CPs are
licensed by case marking, whereas non-factive ones - by T-marking. In the
latter case V raises to C for T-linking. This creates conditions for trace
licensing by the V head. On the contrary, factive CP complements have a
nominal nature and extraction from non T-linked yields that-trace effect.
This analysis is supported by the fact that in "regret" sentences EV2 is
Under negation, the negative operator is in C, which forces V to C
movement and thus the subject trace may be licensed. This movement may be
sometimes overt in English:
(27) Who did John say that never in his life had t been insulted like this?
Part III. The morphosyntax of nominal and verbal constituents
Bracketing Paradoxes Do Not Exist (with Harry van der Hulst and Frans van
The paper considers rules of word formation in English. Two approaches to
the derivation of the noun "ungrammaticality" are possible: (28.a) ("un"
may be attached only to adjectives) and (28.b) (the so-called "level
(28) a. [[[un]grammatical]ity] b. [un[grammatical[ity]]]
Hoekstra et al. distinguish three types of bracketing paradoxes:
(a) unhappier-type, where order "unhappy" - "unhappier" violates the
phonological rule, whereas "happier" - "unhappier" is semantically ill-
formed and is not consistent with the so-called level ordering theory
(inflectional morphology must be attached after the derivational,
compounding must follow affixation, etc.);
(b) modifier scope paradox: "ungrammatical", "model-theoretic",
Dutch "blauwogig" ("blue-eyed"); and
(c) deverbal compounds: "truck-driver".
One of the most interesting questions is thus: is the level ordering
theory on the right track or must it be rejected?
Regarding the (a)-type paradox, they note that it is not a problem at all
and that two possible ways of generating "unhappier" correspond to two
levels of grammar (which may not go hand in hand): phonological and
As for (b), it is argued that some examples clearly force us to abandon
the level ordering hypothesis, while the others was assigned semantic
structure in the wrong manner. In the former case the proposal is that the
words "ungrammatical", "model-theoretic" etc. are formed by the consequent
attachment of the most "native" morphemes to the lexical head ("stratal
constraints" on morpheme combination). In the latter case the traditional
analysis of "blauwogig" as semantically derived from the paraphrase "with
blue eyes" is not valid. Hoekstra et al. criticize the paraphrase method
and point out that not only morphologically complex items may receive
more than one paraphrase, but also such non-derived lexemes
as "strategist", cf.: "a great strategist" etc.
Finally as for (c), there is no reason to believe that in "truck driver"
or "driver of a truck", "driver" receives its theta role from the verb
stem "drive", because in many obviously nominal stems theta role may also
be assigned by the head noun: "book author" / "author of a book" etc.
To summarize, "the category of bracketing paradoxes falls apart into a
number of heterogeneous categories for which separate approaches are
The Nominal Infinitive (with Pim Wehrmann)
This paper concentrates on the infinitival nominalization. As Hoekstra and
Wehrmann observe, there are three types of infinitive constructions in
Dutch that exhibit the external behavior of noun phrases and at the same
time have the verbal core. The first type, (29), allows for both adverbial
and adjectival modification, places their prepositional complements either
before or after the infinitive head and takes the articles.
het schillen van aardappels
the peel-Inf of potatoes
'the peeling of potatoes'
Another type of nominalized infinitive, (30), goes further in getting
verbal properties: it can be preceded by the (in)direct object NPs as well
as by adverbial adjuncts.
het aardappels schillen
the potatoes peel-Inf
the peeling of potatoes'
The third type of nominalizations, (31), which is formed in the absence of
the definite article, seems the most verbal one. It can not have "van"
("of") complements and does not receive adjectival modification. It also
can not take "door" ("by") adjuncts.
Only the latter type of nominalizations contain Vmax (Verb maximal
projection) and have PRO subject position. All the nominalizations in
question inherit their external nominal properties from the nominalizer "-
en", the head of the construction.
Impersonal passives are impossible in both types of nominalizations due to
the following: PRO in "het"-less nominalizations must govern the trace in
the object position, but there are no objects in the active counterpart of
impersonal passives. Thus, no thematic role is associated with PRO.
In "het"-nominalizations with impersonal passives, not only accusative, but
also nominative case assigner is absent: there is no Vmax in this case.
Thus, the passive morphology (see above) argument can not receive any case
At the end of the article some parallels between "het"-less
nominalizations and non-countable nouns are discussed.
Parallels Between Nominal and Verbal Projections
One of the main topics of the paper is the so-called "of"-insertion. What
is not quite clear from the original Chomsky (1970) formulation is
whether "of"-phrase is a PP or NP. Here Hoekstra again examines the
nominalized infinitives in Dutch, which differ from their English
counterparts in the pre-head position of the unmarked direct object (since
Dutch is an OV language).
Hoekstra assumes, following Kayne (1994) that prepositional objects move
from the post-head position to Spec,AgrOP in one type of nominalization
and receive nominal case by virtue of "van" ("of")-insertion in the other
type. But then the question arises: why may only indefinite objects move
to preverbal position in nominalization, whereas in simple clause the
option of object raising is available only for definite NPs?
One more puzzle is how to explain cases where (postoposed) van-NPs may
bind (take scope of) the elements preceding the verb:
het over zichzelf praten van Jan
the about himself talk-Inf of John
'John's talking about himself'
Hoekstra defends Kayne's analysis of DP, according to which DP is an IP or
[D [CP]] where "of" is a C. Under this view Hebrew construct state and
free state constructions can receive strict analysis as in (33.a) and
(33) a. [DP N(i) [IP [DP] I ... [... N(i) ...]]]
b. [DP D [CP [NP(i)] C [IP [DP] I ... [NP(i)]]]]
The same analysis can be adopted to Germanic 's-genitive constructions
and "of"-constructions with the difference that in Hebrew it is the
possessed head N-to-D movement that feeds DP, whereas in Germanic such
movement is the SpecIP-to-SpecDP movement of possessor.
Infinitival nominalizations then have the same [D [IP]]-structure. This
accounts for the backward control mentioned above: V+N phrase complement
of I, which raises to Spec, CP in accordance with (33.b), is lower that
("van")-possessor in the initial structure and thus is c-commanded by the
The analysis in (33) can explain also why only non-specific objects occur
in the pre-infinitival position - this is the result of the whole VP
movement to Spec, CP. On the contrary, definite DPs must scramble out of
the object position to the complement of "van" due to the definiteness
restrictions on object raising.
In the rest of the paper Hoekstra discusses the constraint on
complementation in some syntactic environment, as for instance in (34):
(34) *a proud of his children man
He formulates the following constraint for such cases:
(35) The structure D [CP [XP ... X YP] C [... is ill-formed for certain
choices of X, YP, and, C, if YP is non-null.
This constraint results in the ungrammaticality of constructions like:
(36) *the examining in the yard of the patient
Part IV. Small clauses
Complex Verbs (with Monic Lansu and Marion Westerduin)
Hoekstra et al. start the discussion with the following observation: the
postposition of PPs in Dutch, allowed in other cases ("PP-over-V"), in
resultative constructions is prohibited. But when the verb is prefixed, PP
postposition is again grammatical:
dat ik hem [tot voorzitter] benoem / [tot voorzitter]
that I him to chairman appoint / to chairman
'... that I appoint him chairman.'
Moreover, in these constructions the resultative attributes may be omitted
without changing the meaning of the verb. The proposal here is that in
constructions in questions, PPs are not complements but adjuncts.
The authors argue that in general SCs are complements with a trace in the
subject position that must be governed by the matrix subject or object.
According to Stowell (1981) small clauses are maximal projections of their
heads. The PP in (37) can not be moved out past the verb since it is not a
maximal projection. The whole small clause can not be moved either since SCs
must be canonically governed by the verb.
It is claimed, then, that the only instance of SC constructions where a small
clause lacks the resultative reading are sentences with the stative matrix
(38) *Medusa saw the hero stone / into stone.
Hoekstra et al. observe that some examples of transitive clauses with SCs
become ungrammatical in the absence of SC, (39.a). At the same time they
are perfectly correct if the verb is prefixed, (39.b):
dat Jan zich *(vol) drinkt
that John self full drinks
'... that Jan drinks himself into a stupor.'
dat Jan zich be-drinkt
that John self drinks
'... that Jan drinks himself stupid.'
The main proposal is to treat a verbal prefix as a small clause predicate
with its own argument which surfaces as the matrix object ("zich"). The
structure in (40) thus allows one to explain some properties of "be" / "ver"-
prefixed verbs in Dutch.
(40) dat Jan [SC [NP] be / ver ] V
Under this assumption the difference between Dutch and English can be
accounted for: whereas in Dutch prefixed verbs "with" PPs are adjuncts to
the prefix predicate, in English they are parts of resultative SCs. This
explains the fact that in Dutch but not in English "with"-PPs are
de muur beplaken
the wall plaster
b. stick the wall *(with leaflets)
Small Clauses Everywhere
This article is a part of the book which Hoekstra started to write but
ceased working on long before his death.
In the beginning of the paper he observes three possible analyses of SCs:
the predication analysis (42.a), the complex predicate formation (CPF)
analysis (42.b) and the SC analysis (42.c).
(42) a. We found [NP John](i) [AP guilty](i).
b. We [found guilty] John.
c. We found [SC John guilty].
Based on different properties of the constructions in question (theta-
assignment, word order, PP extraposition in Dutch, and others), Hoekstra
himself defends the SC approach. Following Stowell (1981) and Chomsky
(1981) he defines SCs as the predicates, the extended projection of which
is the AgrP. The subject (of SC) then is defined as follows:
(43) Subject: an agreeing specifier position.
Argumentation for (43) is provided with some examples from the
distribution of the Dutch pronoun "het" and French pseudo-relative clauses.
Another piece of evidence is adduced in favour of the SC analysis. If one
assumes, following Dotjes (1992), that floating quantifiers (FQ) adjoin to
some projection, if this projection contains a trace of the removed NP
modified by this FQ, examples like (44) also support SC analysis:
(44) Il a [XP tous(i) [XP voulu [ les voir t(i) ] ] ].
One of the main contributions to syntactic theory concerns theta-
assignment. Hoekstra reformulates theta-criterion in the following manner.
Theta roles are assigned inside the lexical projection of SC, but every
noun phrase may receive no more than one theta-role in the projection of
its lexical head. Then, the subjects of SCs, which seem to be theta marked
twice (by the SC predicate and by the matrix verb) can be regarded as
(45) X [ NP(i) F [t(i) Y NP ]], where the small clause subject NP(i) is
first theta-marked by the SC predicate Y and then moves to some Spec, FP
to receive theta-role from the matrix verb X (thus the condition on the
theta-assignment inside lexical projections is not violated).
Another serious (but not indisputable) elaboration of syntactic theory is
the proposal to reduce complements and specifiers in favour of adjuncts.
According to Hoekstra, distinction between arguments and adjuncts can be
captured in terms of theta-assignment. Another important difference, the
difference between external and internal arguments, may be reduced to the
aspectual properties of the predicates on the one hand, and to the
configurational distinction on the other. In the latter case Hoekstra
postulates empty predicates for some class of verbs.
Hoekstra also observes that SCs are statives (with a possible empty
head) which must be embedded under some predicates (primarily, non-telic
ones) and denote the final state in the activity, represented by the whole
predication. On other hand, there are also some empty predicates (CAUSE in
(46)) that take the clause introduced by the verb as their argument:
Jan bevuilt zijn kamer
John BE-dirty-s his room
Jan CAUSE [SC [zijn kamer] be-[vuil] ]
More generally, Hoekstra argues that the SC schemata, (47), underlies many
types of constructions, for instance, constructions with light verbs,
serial verbs, double objects, verb particles and others.
(47) NP Pr [ NP Pr NP]
Such an analysis has a very important theoretical advantage - it helps to
save binary branching in the case of double objects and similar constructions.
One of the most characteristic features of Hoekstra's approach to language
is that his researches were always on the front line of linguistic theory.
Thus, as has become clearly evident from this volume, the Lexical
Decomposition methods, split VP hypothesis, Abney's hypothesis, an
approach in the spirit of Distributed Morphology, and other remarkable
proposals were either adopted by Hoekstra just after their appearance in
the linguistics literature or even anticipated by him.
Teun Hoekstra's viewpoints on this or that linguistic problem may seem
quite unusual at first sight. But all his ideas, often not commonly
accepted, constitute a very consistent way of thinking, which he tried to
lead through the wide range of linguistic material. His proposal of T-
chains, views on passive morphology as anaphoric pronoun with argument
function, denial of the existence of verbs as a lexical category,
preposition / empty predicate incorporation into the lexical verb, and
the "everywhere" nature of SCs may be cited among ideas that Hoekstra
proposed and tried to develop in his works.
An interesting question that comes into one's mind when reading the book
is: what would linguistic theory be if Hoekstra's proposals
constituted the basis of syntactic theory? Would we have a more or less
powerful apparatus at hand than we do now? What will the consequences
be for other areas of morphosyntax not directly addressed by Hoekstra?
However, this collection of Hoekstra's writing seems not only interesting,
but also useful reading for contemporary researchers, since both the
problems and the ideas discussed in this volume are very far from being
Chomsky, Noam. 1970. Remarks on nominalization. In Readings in English
transformational grammar, Roderick Jackobs and Peter Rosenbaum (eds.), 184-
221. Waltham, MA: Ginn.
Chomsky, Noam. 1981. Lectures on Government and Binding. Dodrecht: Foris
Doetjes, Jenny. 1992. Rightward floating quantifiers float to the left.
The Linguistic Review 9 (4): 313-332.
Hale, Kenneth and Samuel Jay Keyser. 1993. On argument structure and the
lexical expression of syntactic relations. In The view from building 20:
essays in linguistics in honor of Sylvain Bromberger, Kenneth Hale and
Samuel Jay Keyser (eds.), 53-109. Cambridge, MA: MIT Press.
Kayne, Richard. 1985. L'accord du participe passé en français et en
italien. Modèles Linguistiques 7 (1): 73-89.
Kayne, Richard. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press.
Pollock, Jean-Yves. 1989. Verb movement, universal grammar and the
structure of IP. Linguistic Inquiry 20 (3): 365-424.
Roberts, Ian. 1985. Absorption parameters and the passive in UG. Paper
presented at GLOW 8, Brussels. Abstract in GLOW Newsletter 14: 71-73.
Stowell, Tim. 1981. Origins of phrase structure. Doctoral dissertation,
ABOUT THE REVIEWER
Pavel Grashchenkov has graduated from the Department of Theoretical and
Applied Linguistics of Moscow State University. His diploma "Typology of
Numeral Phrase" deals with the morphosyntax of the noun phrases with
numerals in the languages of the world. Now he is finishing his
thesis "Syntax and Typology of Genitives", in which some problems
concerning the position and the status of genitive noun phrases are
discussed. His academic interests include: Morphosyntax of DP - syntax and
typology of PossP, NumP and other DP-internal projections, structure of
APs; Case Theory - inherent vs. structural cases, grammatical cases in
ergative languages, genitive case assignment; Typology - Caucasian,
Turkic, Slavic, Finno-Ugric languages; Part-of-Speech Systems - languages
without lexical classes, non-universal lexical classes.