REVIEW ARTICLE

CATEGORIZATION AND LICENSING OF

SUBORDINATING CONJUNCTIONS

YOSHIO NAKAMURA
Yokohama National University*

The Syntax of Subordination, by Dagmar Haumann, Niemeyer, Tubing-

en, 1997, viii+278pp.

Keywords: subordinating conjunctions, complementizers, prepositions,

functional architecture, feature specification

1. Introduction

The question of how subordinating elements are syntactically related

to the matrix clause has been and still is a matter of debate. The Syn-
tax of Subordination provides a thoroughly worked out syntactic analy-
sis of the issue. Focusing on the discussion of categoryhood and
licensing of subordinate conjunctions and subordinate clauses, this book
offers a novel explanation as to how to handle the syntax of subordina-
tion under the recent generative linguistic theory.
The Syntax of Subordination is organized into three parts. Part I
(Early Generative Grammar, the Rise of Category C and the Categoriza-
tion of Subordinating Conjunctions) reviews the conception of phrase
structure in early generative grammar and past analyses of subordina-
tion. In chapter 1 (Subordination in Early Generative Grammar) the
distribution of noun clauses and adverbial clauses is discussed. In
chapter 2 (Complementizers) two traditional analyses of complementiz-
ers (the transformational approach and the phrase structure rule analy-
sis) are reviewed. In chapter 3 (Subordinating Conjunctions-A

*I would like to thank two anonymous EL reviewers for their comments and

suggestions on both content and exposition. All remaining inadequacies are my

own.

English Linguistics 18: 2 (2001) 593-618 -593-

594 ENGLISH LINGUISTICS, VOL. 18, NO. 2 (2001)

Question of Categorization) the categorization options for subordinate

conjunctions are discussed.
Part II (Recent Conceptions of Phrase Structure) summarizes the con-
ceptions of phrase structure in the Principles and Parameters Theory
and the Minimalist Program. In chapter 4 (Phrase Structure) the dis-
tinction between lexical and functional heads is discussed. In chapter
5 (Clause Structure) clause structure is examined in terms of the "split-
Infl" hypothesis.
Part III (Lexical Properties of Complementizers, Prepositions and
Subordinating Conjunctions) develops a theory of subordinate conjunc-
tions. In chapter 6 (Lexical Properties of Complementizers and Prop-
erties of Category C) it is argued that complementizers make up a func-
tionally defined category, isolating the lexical properties of lexical ele-
ments in the C position. In chapter 7 (Lexical Properties of Preposi-
tions) it is argued that PP is dominated by functional architecture. In
chapter 8 (Subordinating Conjunctions Straddle the Dividing Line be-
tween the Lexical and the Functional Universe) a third category (Sub-
con) is introduced as a conflating category, unifying properties of lexi-
cal and functional heads. In the concluding chapter (chapter 9) the
derivation of the subordinate-matrix order is discussed.
Part I and Part II can be seen as a survey of past analyses and the
current state of linguistic theory of subordinating elements. These
parts set a stage for part III. In the following three sections, I outline
Haumann's arguments concerning complementizers and their feature
specifications (section 2), the lexical prepositions that introduce tempo-
ral adverbial clauses and their extended projections (section 3), and
subordinating conjunctions as a distinct syntactic category which con-
flates properties of both the functional head C and the lexical head P
(section 4). In section 5, I discuss some problems and consequences of
Haumann's analysis. Section 6 is the conclusion.

2. Properties of Complementizers
Since Bresnan (1970, 1972, 1973), complementizers have been analy-
zed as being a syntactic category. The category C is not only a label
but also a functional head that projects CP. Assuming that com-
plementizers serve to introduce argument clauses, Chapter 6 is devoted
to examining lexical properties of complementizers and properties of
 CATEGORIZATION AND LICENSING OF SUBORDINATING CONJUNCTIONS 595

C.1 Haumann analyzes that and if as complementizers. On the other

hand, Haumann analyzes whether as an interrogative specifier. For is
argued to be a Case-assigner inserted into C. Trying to isolate the lex-
ical properties of the items base-generated in C, complementizers and
the structure of the clause they head are discussed in chapter 6.
Let us first see Haumann's analysis of that complementizers. They
are divided into two groups: the indicative complementizer thatind (icative)
and the subjunctive complementizer thatsubj (unctive). CPs headed by
thatind are of two kinds: factive and non-factive. Factive CPs induce
definiteness in the sense that their distribution is restricted to the inter-
nal argument position of "predicates which assert or presuppose truth
value, or which imply knowledge or certainty" (Bresnan (1972: 65)).
Melvold (1991) argues that thatind in factive CPs is analyzed as bearing
the feature [+def (initeness)] which licenses the iota operator in a spec-
head-agreement manner. Basically following Melvold (1991), Haumann
argues that there are two thatind-complementizers, one that heads fac-
tive CPs and one that heads non-factive CPs. In factive CPs, thatind is
specified as [+def]. It licenses an operator in the specCP, marking it
as [+def]. On the other hand, in non-factive CPs, thatindis specified as
[-def], so that its specifier is also marked as [-def]. Note that, con-
trary to Melvold (1991), Haumann does not assume that that introduc-
ing non-factive CPs is semantically inert and inserted at PF.
For the agreement between a [+def]-head and its specifier, Haumann
proposes the following formulation:
(1) a. A [+def]-Operator must be in a Spec-head configuration
with an X0[+def]
b. An X0[+def]must be in a Spec-head configuration with a
[+def]-Operator. (p. 157)
Since a [+def] operator occupies the specifier position of factive CPs
according to (1), specCP is unavailable as an intermediate landing site
for long wh-extraction. Thus, argument extraction out of factive CPs
causes a weak island effect and adjunct extraction yields a strong viola-
tion:
(2) a. ?Whoi did Fred confess that he fired ti?

1 Following Riemsdijk (1984), Haumann assumes that clauses with a lexical com-
plementizer are CPs. On the other hand, Haumann rejects the assumption made
by Abney (1987) and Fukui and Speas (1986) that the relation between C and
AgrsP is characterized by f(unctional)-selection.
596 ENGLISH LINGUISTICS, VOL. 18, NO. 2 (2001)

b. *Howi did Bill reveal that Ann solved the problem ti?
(p. 151)
On the other hand, extraction out of CPs headed by non-factive thatina
is possible:
(3) a. Whoi did John believe that Susan invited ti?
b. Howi did Bill believe that Ann solved the problem ti?
(ibid.)
Since the specifier position of non-factive thatind does not contain an
operator, it is available for wh-movement.
Furthermore, factive complements differ from non-factive arguments
in that they may be wh-initial:
(4) a. Fred figured out whoi Peter eloped with ti.
b. The scientists realized immediately whati they had dis-
covered ti. (p. 155)
(5) a. *Fred insisted whoi Peter eloped with ti.
b. *The scientists claim whati they had discovered ti. (ibid.)
For these examples, Haumann assumes that (non-) interrogativity is rep-

(1991). Although all wh-initial CPs are marked as [+wh], only those
which are marked as [+Q] are interrogative and thus indefinite. Since
the argument CPs of the factive verbs are not interrogative in (4), they
are interpreted as [-Q]. Such wh-initial CPs, being [-Q], count as
definite. The definiteness of the CPs is accounted for by specifying C
as [+wh] and [-Q]. According to the requirements stated in (1), the
wh-phrases in the specifier positions of the factive complements are
specified as [-Q] and [+def]. On the other hand, [-def] CP-argu-
ments of non-factive predicates cannot be wh-initial, as shown in (5).
Since the governing verbs are not interrogative, they cannot license wh-
initial complements.
Although Haumann recognizes two thattnd-complementizers (one that
introduces [+def] CPs and one that introduces [-def] CPs), thatind-com-
plementizers are always finite. Haumann assumes that (non-) flniteness

analysis, Agrs-to-C movement is deferred until LF unless there is some

strong feature in C. Overt Agrs-to-C movement is assumed to be
forced by strong features such as [+Q] in matrix interrogatives, [+Agr]
in V2 contexts, [+subj (unctive)] in V2 argument clauses in German,
and [+cond (itional)] in counterfactual conditionals in English and Ger-
man. Relevant cases are exemplified below:
 CATEGORIZATION AND LICENSING OF SUBORDINATING CONJUNCTIONS 597

(6) a. Will she fix the car?

b. Sie hat das Auto nicht repariert.
she has the car not fixed
She hasn't fixed the car.'
'
c. Sie sagte, sie werde das Auto nicht reparieren.
she says she Willsubj the car not fixed
'She said she would not fix the car.'
d. Had she repaired the car, he might still be alive.
(pp. 160-161)
In the examples given in (6), overt Agrs-to-C movement has taken
place as a consequence of C specified by strong features such as [+Q],
[+Agr], [+subj], and [+cond]. In (6c), for example, the subjunctive
argument clause is characterized by the feature [+subj]. Being strong,
the feature [+subj] needs to be lexicalized in overt syntax.
Notice here that, in German, lexicalization of [+subj] is executed
either by the base-generation of dasssubj or by Agrs-to C movement:
(7) a. Sie sagte, sie werde das Auto nicht reparieren.
she says she Willsubj the car not fix
b. Sie sagte, dass sie das Auto nicht reparieren werde.
she says that she the car not fix Willsubj
(p. 161)
In English, C[+subj] selected by verbs such as require must be lexicalized
as the base-generated thatsubj:
(8) a. It is required that he shave before noon.
b. The situation demanded that the dishes be done im-
mediately. (p. 164)
Recall that under Haumann's analysis subjunctive CPs are taken to be
finite. Assuming, following Roberts (1993), that every verb must be
PF-identified by being associated with some inflectional feature of a
governing head, Haumann argues that T in subjunctive CPs may be
occupied by a modal null element.2 Thus, lexical materials such as the
aspectual auxiliary have cannot occur in the T position:
(9) a. *I suggest that he have not left before noon.
b. *I suggest that he be not there by 8.
c. I suggest that he not have left before noon.

2 In British English, should is inserted in T of subjunctive CPs.

598 ENGLISH LINGUISTICS, VOL. 18, NO. 2 (2001)

d. I suggest that he not be there by 8. (p. 167)

In negative preposing cases, the null modal 0 (or the overt modal
should) undergoes overt movement over the subject:
(10) a. ... that under no circumstances should he bring out the
trash.
b. ... that under no circumstances 0 he bring out the trash.
(adapted from p. 171)
Under Haumann's analysis, if is also analyzed as a complementizer.
Argument CPs headed by if are finite and restricted to the internal
argument position of interrogative heads.
(11) a. Joe asked if Tom knew about the gossip.
b. *Joe asked if Tom to know about the gossip.
c. *Joe said if Tom know about the gossip. (pp. 171-172)
If is thus assumed to be marked as [-fin] and [+Q]. The specifier
position of the CP if heads is also marked as [+Q]:
(12)

(p. 172)
Since CPs headed by if host an empty operator as depicted in (12),
they show island effects as factive CPs do
(13) a. ?Whati did you ask [if Tom knew ti]?
b. *Howi did, you wonder [if Tom made it home ti]? (p. 173)
Argument CPs with a wh-element marked as [+Q] in its specifier are
taken as indefinite arguments in the sense of Bresnan (1972) and Baker
(1970). In this regard, CPs headed by if resemble non-factive CPs in
that both show indefiniteness.
Like CPs headed by if, argument CPs introduced by whether, being
specified as interrogative, appear in the internal argument position of
an interrogative head:
(14) a. I am asking whether you will accompany me.
b. *I think whether you will accompany me. (p. 174)
Argument CPs introduced by whether are differentiated from argument
CPs headed by if/thatlndisubjin that they can be either finite or non-
finite
(15) 1 wonder whether to go home right now. (ibid.)
 CATEGORIZATION AND LICENSING OF SUBORDINATING CONJUNCTIONS 599

Assuming that whether is the lexical counterpart of the empty inter-

rogative operator, Haumann argues that whether occupies specCP, lex-
icalizing [+Q].3 Haumann also assumes that whether, being a specCP-
element, is not specified for the head feature
specCP is analyzed as being in a specifier-head-agreement relation with
an empty [+Q]-complementizer.
In the end of chapter 6, Haumann examines the properties of for.
For takes non-finite declarative clauses with a lexical subject:
(16) a. We hope for him to succeed.
b. *We hope for PRO to succeed. (p. 178)
Governing verbs such as try, convince, hope, want, etc. select com-
plementizers associated with future tense. The [+fut] feature in C
selects a non-finite T, a T head which is lexically filled by to.
Haumann takes for as a (de-) prepositional complementizer that func-
tions as a structural Case-assizner for a lexical subject:
(17)

(p. 187)

3 As supporting evidence for the claim that whether is a specifier, Haumann cites
the observation of Lightfoot (1979) that whether and that, a genuine C-head, could
cooccur in earlier stages of English.
600 ENGLISH LINGUISTICS, VOL. 18, NO. 2 (2001)

Under this analysis, for is not a complementizer but a Case-assigner,

and its presence has nothing to do with selection. Since neither [-fin]
nor [+fut] are taken to be strong features, for cannot be regarded as a
lexicalization of such features. The head of the CPs of the form
"PRO -to-...-VP" is a non-lexical counterpart of for, characterized by
identical features [-fin] and [+fut]. The CPs of the form "for-DP-to-
-VP" and the CPs of the form "PRO-to-...-VP" are both ...marked as
[-def].

3. Lexical Prepositions and the Functional Architecture

Whether prepositions such as before, after, and since with sentential
complements are lexical or functional has been a long-standing ques-
tion. In the recent literature, there have been two major analyses.
Webelhuth (1992) analyzes them as prepositions taking IP comple-
ments. Huang (1982) and Dubinsky and Williams (1995) take them as
complementizers. Chapter 7 is devoted to examining the lexical prop-
erties of prepositions. In this section, I review Haumann's new analy-
sis of the prepositional heads that take sentential complements forming
temporal adverbial clauses. Prepositions of the subclass are termed
lexical prepositions.4 It is argued that the prepositional elements show
thematic properties. Under Haumann's analysis, "matrix clauses" in
the traditional sense are analyzed as the external arguments licensed by
head-specifier relation within PP. Furthermore, Haumann develops a
theory of the functional structure dominating PPs.
The lexical prepositions after, before, since, until, when, and while
are associated with an argument structure containing three arguments:
an external, an internal, and a referential argument (a spatial <s> or a
temporal argument <t>). (In) transitivity depends on whether an in-
ternal argument is encoded or not.
In addition to the argument structures, prepositions are lexically
specified for Case-marking properties. After, before, since, and until in

4 Prepositional elements that are not associated with a lexical argument structure
are called Caseprepositions (or grammatical(ized) prepositions):
(i) He is good at tennis. (p. 191)
 CATEGORIZATION AND LICENSING OF SUBORDINATING CONJUNCTIONS 601

(18) take a nominal internal argument which must be assigned a

structural Case
(18) a. We planned to meet after the conference.
b. I didn't see him before the conference.
c. I haven't seen him since the conference.
d. We couldn't get hold of him until the end of the confer-
ence. (p. 196)
These prepositions have the following specification:
(19) {<1,2,t>,[+CAS],[+S]} (ibid.)
Here, <1> designates the external argument, <2> the internal argu-

herent Case ([-S]). The following specification (20) accounts for af-
ter, before, since, until, when, and while in (21):
(20) {1<1,2,t>,[-CAS]} (ibid.)
(21) a. We planned to meet after the conference was over.
b. I didn't see him before the conference started.
c. I haven't seen him since the conference started.
d. We couldn't get hold of him until the conference was
over.
e. I met him when I was in Paris.
f. We talked about Jill while the kids were out. (ibid.)
The internal argument is discharged by theta-marking.5 In the follow-
ing structure, the internal argument of the preposition before (<2>) is
coindexed with the referential argument of the semantic head of the
complement (N in (22a) and V in (22b)):6

5 Theta marking is defined as follows;

(i) A lexical head L theta-marks a phrase FP iff:
a. L governs FP, and
b. the referential argument of the semantic head of FP is coindexed
with a thematic argument of L. (p. 194)
6 In (22) and (24), e stands for a referential argument e (vent).
602 ENGLISH LINGUISTICS, VOL. 18, NO. 2 (2001)

(22)

a.before the conference

<1, 2i,t> <Ri>
b. before the conference started
<1, 2i,t> <ei> (p. 194)
Haumann further argues that the external argument is discharged by
theta-marking a maximal projection, specPP:
(23)

<..., ref. arg. j><1j, 2i,t><..., ref. arg.i> (p. 199)

Under this analysis, the matrix clause under the traditional analyses is
taken as specPP licensed within the lexical projection of P, and sub-
ordinate clauses are taken as the prepositional predicate.
(24)

We had a party after she left

They all left before the fireworks started
I haven't seen him since the conference
You will have to wait until noon
We met when I was in Paris
We planned to meet while the kids were out
<..., ej><1j, 2i,t><..., ref. arg. i> (ibid.)
Lexically defined properties of P require and license the presence of
a functional structure above PP. There are at most three functional
projections dominating PP as depicted below:
 CATEGORIZATION AND LICENSING OF SUBORDINATING CONJUNCTIONS 603

(25)

(p. 233)
The presence of P-AgroP is triggered by prepositions that are associ-
ated with an argument structure containing an internal argument and
are specified as structural Case-assigners:
(26) {<1, 2, t>,[+CAS],[+S]} (p. 196)
The prepositional/postpositional order in (27) is accounted for by
assuming overt/covert movement of the direct object of P into the spec-
ifier position of P-AgroP:
(27) a. They met before the conference.
b. They met three weeks ago. (p. 205)
(28) a.
604 ENGLISH LINGUISTICS, VOL. 18, NO. 2 (2001)

b.

(adapted from p. 206)

Haumann assumes that [+S]-feature of ago is strong so that it triggers
the overt movement to P-Agro and specP-AgroP. On the other hand,
in cases such as (27a), movement of P into P-AgroP and movement of
DP into specP-AgroP are deferred until LF.
The presence of P-DegP is triggered and licensed by the fact that the
argument structure of P contains a referential argument.7 Haumann
further argues that modificational elements such as right, long, and two
hours in (29) are base-generated in specP-DegP as represented in (30):
(29) a. The dizzy turns started right/long/two hours after the
conference started.
b. The dizzy turns started right/longl two hours before the
conference started. (p. 218)

7 The referential argument is discharged by being theta-bound by a functional

head, that is, by being coindexed with the functional head P-Deg.
 CATEGORIZATION AND LICENSING OF SUBORDINATING CONJUNCTIONS 605

(30)

right the dizzy after the conference started

two hours turns started <1, 2, t>
[+time]
[-perf]
[-simult] (p. 221)
Here, (non) simultaneity is specified as

ificational elements such as right and two hours are assumed to have
lexical specifications that enter into a specifier-head agreement relation
with P-Deg.8 In LF, the head P moves into P-Deg to have its [+time]-
feature checked.
The subject of P (the matrix clause under the traditional sense)
moves to some higher position preceding both P-AgroP and P-DegP:
(31) a. [He left]i[P-DegP two hours [P-AgroP [pp ti [p' after the party
started]]]]
b. *[P-DegP Two hours [P-AgroP [PP he left [P' after the party
started]]]]
(adapted from p. 227)
Haumann argues that the subject undergoes overt movement to the

8 Elements such as right and two hours are specified for their ability to measure
(temporal) distances in terms of degrees. Right is marked as degree zero ([deg-0])
and measure phrases such as two hours are marked as degree n ([deg-n]).
606 ENGLISH LINGUISTICS, VOL. 18, NO. 2 (2001)

(32)

he lefti two hours ti after the party started

(adapted from p. 229)
For the question of why sentential elements should move overtly,
Haumann suggests that sentence-type features are strong and trigger

here that, contrary to the sentential internal argument of P (AgrsP),

the sentential external arguments vary in sentence-type (declarative, in-
terrogative, etc.) and they can be realized as AgrsP, CP, or TopP, as
the following examples show:

finished the dishes]]. (adapted from p. 231)

and hosts the subject of P.

For the following examples, Haumann argues that the matrix gov-

P, which bears sentence type information:9

9 Haumann takes selection as selection of either the highest specifier or the high-
 CATEGORIZATION AND LICENSING OF SUBORDINATING CONJUNCTIONS 607

party started]].

party started]]. (adapted from p. 231)

4. Subordinating Conjunctions
Chapter 8 develops a theory of subordinating conjunctions such as
although, because, causal for, if, in case, in order, causal since, unless,
conditional whether, and adversative while. Haumann argues that
these elements cannot be unambiguously attributed to either com-
plementizer or preposition. Haumann proposes subordinating con-
junctions should be analyzed as a distinct syntactic category Subcon,
which conflates properties of both the functional head C and the lexical
head P.

4.1. Functional Properties of Subcon

In the case of the category C, a functional head, the relation between
a C head and its sentential complement is established by checking fea-

ecuted by covert movement of a verbal material into a C head. In

chapter 8, Haumann first examines the relation between subordinating
conjunctions and their structural complement.
Subordinating conjunctions take finite sentential complements, as
shown in (35). Haumann assumes that the feature [+fin] determines
the relation between Subcon and the sentential complement in these
cases: 10
(35) a. Although he had no driver's license, Tom drove us
home.
b. He didn't show up at the conference because he hadn't
read the abstracts.
c. She didn't come to the party for she had papers to
grade.

est head of the extended projection.

10Haumann argues in chapter 9 that the subordinate-matrixorder in cases such
as (35a), (35d), (35f), (35h), and (35i) is derived as a result of leftward movement
of the subordinate clause.
608 ENGLISH LINGUISTICS, VOL. 18, NO. 2 (2001)

d. If she had not forgotten to grade the papers, she would

have come to the party.
e. I had to watch where I put my feet in case I left.
f. Since you are a man you are supposed to know about
this.
g. We won't get the train unless we leave at 5 a.m.
h. Whether you like or not, I will do it.
i. While John refuses to take up linguistics himself, he
talked Tom into it. (p. 238)
In order is specified as [+fut] and takes a CP complement marked as
{[+fut], [-fin]}(as in (36)) or {[+fut], [-fin], [+subj]} (as in (37)):
(36) a. She came home early in order to get some sleep.
b. She typed the paper in order for him to get some sleep.
(p. 239)
(37) a. She typed the paper in order that he might get some
sleep.
b. She came home early in order that she get some sleep.
(ibid.)
Although, whether, and while can be marked as {[-fut], [-fin]}, allow-
ing gerundive sentential arguments:
(38) a. Although playing with the wind in the first half they
were never able to control the play. (p. 238)
b. Whether living in London or not, John employed him-
self.
c. While refusing to take up linguistics himself, John talked
Tom into it. (p. 239)
The structural complement of although, whether, if, and unless marked
as [-fut] is realized as a small clause:11
(39) a. Although drunk, he made it home alive.
b. Whether in a bad mood or not, don't call me.
c. While drunk himself, John tried to prevent Tom from
drinking too much.
d. If tired, she will not come to the party.

11 Although Haumann (p . 243 (15)) marks the ability of although, if, unless, and
whether to take a small clause complement, while also takes a small clause comple-
ment as (39c) shows.
 CATEGORIZATION AND LICENSING OF SUBORDINATING CONJUNCTIONS 609

e. Unless in hospital, you have no opportunity to see your

doctor twice a day. (p. 241)
The relation between Subcon and its complement is assumed to be
based on feature checking. For example, Subcons such as although,
because, for, if, in case, since, unless, whether, and while and their
finite sentential complement (AgrsP) are related to the finite features of
the embedded verb. The relation between in order and its CP comple-
ment (finite or non-finite) is based on the feature [+fut] involved in
both Subcon and C.
Haumann takes conditional if to be a prime example of Subcon as a
functional head within the extended verbal projection. It is argued
that the feature [+cond] is involved in examples like the following:
(40) a. If she had repaired the car, he might still be alive.
b. If she were to compromise, this matter would be easier
to settle. (p. 244)
(41) a. Had she repaired the car, he might still be alive.
b. Were she to compromise, this matter would be easier to
settle. (ibid.)
[+cond], being strong, is lexicalized by the base-generation of if or
overt head movement of a finite verb bearing [+cond]-specification:
(42)

lf [+cond] she had [+cond] repaired the car

if [+condl she were [+cond] to compromise
had [+cond]i she ti repaired the car
were [+cond]
i she ti to compromise (ibid.)

4.2. Lexical Properties of Subcon

The relation between Subcon and its sentential complement is re-
610 ENGLISH LINGUISTICS, VOL. 18, NO. 2 (2001)

duced to the checking relation based on the feature of the embedded

verb (or T). In this sense, the projection of subordinating conjunc-
tions is an extended projection of the embedded verb. Put differently,
the relation is not based on thematic properties. On the other hand,
the relation between P and AgrsP is thematic in that P theta-marks and
licenses AgrsP as its internal argument.12
Subcons, however, share some properties with prepositional subordi-
nators. The projections of Subcons are barred from argument posi-
tions within the lexical projection of a head. Furthermore, Subcons
show the lexically defined property of having an external argument:
(43)

<..., ref. arg.j><1j> (p. 252)

Under this analysis, the matrix clause is taken as a base-generated A-
specifier within the projection of Subcon. Haumann further argues
that SubconP is dominated by the functional
on the basis of the lexical part of Subcon. Since the argument struc-
ture of Subcon contains an external argument, SubconP functions as a
predicate:

12 Assuming condition (i) proposed in chapter 7, Haumann observes examples

such as (ii) as evidence for the view that Subcon does not theta-mark its sentential
complement:
(i) The highest specifier in a theta-marked extended projection must be an
A-specifier, unless required otherwise by selectional properties. (p. 213)
(ii) To his surprise he has discovered that although [never before in his life]
had he tried to play tennis right-handed he plays quite well. (p. 248)
 CATEGORIZATION AND LICENSING OF SUBORDINATING CONJUNCTIONS 611

(44)

She tried to appear normali ti although she did not feel well
We refused their proposali ti beause we feared an uproar
She came home earlyi ti in order to get some sleep
(p. 255)

In the following example, the matrix clause has undergone movement

over the focus particle even adjoined to SubconP:

[Subcon'
although [AgrsPpressure was quite hard]]]]]]
(adapted from p. 256)
As is the case with the projection of P, the projection of Subcon is a
predicate of the event referred to by the "matrix" clause, and the spec-

ment of P/Subcon.14

14 In chapter 9, Haumann touches on the derivation of the "subordinate-matrix"

order in (i):
(i) a. After she left, we had a party.
b. Although she did not feel well, she tried to appear normal. (p. 265)
Though detailed mechanics are left unsettled, Haumann assumes that the
"subordinate -matrix" order is derived by leftward movement of P-DegP/SubconP as
depicted below:

ti]]]]

In line with Takano's (1995) analysis of predicate fronting, it is assumed that the
preposed subordinate clauses, being predicate under Haumann's analysis, are inter-
preted in their base positions at LF.
612 ENGLISH LINGUISTICS, VOL. 18, NO. 2 (2001)

5. Some Problems and Consequences

Haumann's novel analysis of subordination gives rise to several prob-

lems and consequences. In 5.1., I take up some problems of
Haumann's proposals concerning that-complementizer shown in section

In 5.2., I discuss some consequences of Haumann's analysis of preposi-

tional subordinators outlined in section 3.

5.1. Problems
Let us first consider that-less cases in non-factive complements.
Since Kiparsky and Kiparsky's (1970) work on factive predicates, the
correlation between the semantic properties of factive complements and
their syntactic behavior has been discussed by a number of researchers.
Hegarty (1991) argues that factive predicates choose a complementizer
that has the ability of
verb. Melvold (1991) argues that the event argument of the embedded
verb is bound by an existential quantifier bound by an iota operator in
specCP. Haumann's analysis of factive CPs is close in spirit to the
analyses proposed by Hegarty (1991) and Melvold (1991).
However, Haumann's analysis crucially differs from such previous
analyses as to how to handle that introducing the non-factive CPs.
Hegarty (1991) assumes that that of non-factive complements has no
semantic role. Similarly, Melvold (1991) assumes that thatind in non-
factive CPs is semantically inert and inserted at PF. On the other
hand, Haumann assumes that the head of non-factive CPs is specified
as [-def]. If this is the case, a question arises here: Why that spec-
ified as [-def] can be deleted in examples such as the following?
(46) He believes (that) the world is flat. (p. 151)
A possible answer is to assume that there exists a null complementizer
(0that) in that-less cases. Since the embedded clause is not definite in
the sense of Haumann, the head 0thatmust be specified as [-def]. It is
then necessary to examine the properties of 0that,though the issue is left
untouched in Haumann's analysis.
Whether finite subordinate clauses which lack an overt com-
plementizer are CPs or IP complements has been a long-standing issue.
For IP-hypothesis, see Doherty (1997). Ormazabal (1995) argues con-
vincingly that both finite and non-finite propositional clauses are CPs.
Under the CP-hypothesis, that-less CPs are headed by a null C. He
 CATEGORIZATION AND LICENSING OF SUBORDINATING CONJUNCTIONS 613

further argues that empty complementizers such as 0thatshould be taken

as affixes. Under Ormazabal' analysis, the following contrast is
accounted for in a different way from Haumann's analysis:15
(47) a. Mary pointed out that [Sue wasn't there].
b. *Mary pointed out [0that[Sue wasn't there].
(Ormazabal (1995: 256))
Ormazabal argues that the main syntactic difference between factive
and non-factive CPs should not be attributed to the nature of their
heads. According to Ormazabal (1995), the difference lies in the posi-
tion in which they are located. He proposes that non-factive CPs re-
main internal to VP at LF while factive CPs are moved out of the c-
command domain of the factive verb that selects them. Being 0-affix,
the null head in (47b) must be incorporated into the matrix verb. At
LF, the complement clause moves into a position higher than the ma-
trix VP:
(48) ... [CPt0thatSue wasn't there] ... [VP ... [V' 0that-VtCP]]...
(adapted from Ormazabal (1995: 263))
The resulting representation is excluded as a form of the Proper Bind-
ing Condition. For detailed discussion, see Ormazabal (1995).
Notice that Ormazabal's analysis is crucially based on the assumption
that factive that and propositional (i.e. non-factive) that belong to the
same semantic type. Much further work is awaited on the nature of
that.
Let us now turn to the question concerning movement of "matrix"

the sentential external argument originating in specSubconP can be

realized as AgrsP, CP, or Tope:
(49) (=(33))
a. He left two hours after the party started.

15 For the obligatoriness of an overt complementizer in factive CPs, Haumann

observes the following examples:
(i) a. *She regrets she had not accepted the offer.
b. *She realized she had run out of coffee.
c. *He pointed out movement out of adjuncts is impossible. (p. 152)
Also, Haumann notes (p. 152, note 10) that judgments vary among speakers for (ia)
and (ib).
614 ENGLISH LINGUISTICS, VOL. 18, NO.2 (2001)

b. Who did he call before the party started?

c. Yesterday, his friends arrived before he had finished the
dishes.
(50) a. We refused their proposal only because we feared an up-
roar.
b. Why did she try to appear normal, although she did not
feel well?
c. This offer, he won't accept it unless you talk him into it.
(p. 259)
Under Haumann's analysis, the "matrix" clause has been moved into

ever, a question arises. Haumann assumes that AgrsP undergoes overt

movement in these cases:

started]]]]]

because [AgrsPwe feared an uproar]]]]]

It has been widely assumed that AgrsP (IP/TP) cannot be a target of
overt movement operation. Some explanation would be called for if
we try to maintain the analysis given in (51). The difficulty can be cir-
cumvented by assuming that the external sentential argument is CP
headed by 0that. We then face a question discussed above concerning
that-less cases in non-factive CPs.
Not only the categorial status of the moved clauses but also the
movement hypothesis is dubious. Consider the following examples:
(52) a. Whoi do you think that [[he called ti] before the party
started].
b. Which paperi do you think that [[she typed ti] in order
that he might get some sleep]?
Under Haumann's analysis, these examples are assigned the following
representations:

before the party started]]]]

[SubconP tj [Subcon' in order that he might get some sleep]]]]

In (53), the external sentential arguments (he called ti in (53a) and she

traction out of the phrases that have undergone overt movement usual-
 CATEGORIZATION AND LICENSING OF SUBORDINATING CONJUNCTIONS 615

ly causes deviancy:16
(54) a. ?*whoi did [IP [a story about ti]j [I'Infl [VP tj [V' amuse
you]]]]? (adapted from Lasnik and Saito (1992: 42))
b. ??whoi do you wonder [CP [which picture of ti]j [C' C [IP
Mary bought tj]]]
(adapted from Lasnik and Saito (1992: 102))
Haumann's analysis would predict that the examples in (53) show deg-
radation, contrary to the fact.

5.2. Some Consequences

Under Haumann's analysis, what we refer to as a subordinate clause
traditionally is the prepositional predicate, and what is being modified
under the standard analyses is taken as the external argument of the
prepositional head. In this sense, "subordinate clauses" are no longer
adjuncts structurally. This analysis seems to be borne out by the facts
concerning extraction. As Lasnik and Saito (1992) observe, extraction
out of the sentential complements of prepositional elements such as be-
fore, after, and since results in mild deviancy:
(55) ??Which linguisti did you write your thesis [after you consulted
ti]? (adapted from Lasnik and Saito (1992: 91))
The example in (55) is better than the example in the following:
(56) *Wherej did you see the book [whichi John put is tj]?
(adapted from Lasnik and Saito (1992: 71))
The difference in acceptability between (55) and (56) follows if we
assume that the "subordinate clause" in (55) is not a pure adjunct.
Haumann assumes that adjuncts in general are predicates of referen-
tial arguments. In (57), for example, the PP after he was administered
the poison is predicated of the event the dizzy turns started:
(57) The dizzy turns started after he was administered the poison.
Under the analysis Haumann develops in chapter 7, what is moved to
be licensed is the external sentential argument (the matrix clause in the
traditional sense). This analysis extends to PPs whose head does not
take an internal argument:
(58) I haven't seen him before. (p. 196)

16 I have no explicit explanation of the mild deviancy of (54b).

616 ENGLISH LINGUISTICS, VOL. 18, NO.2 (2001)

The example in (58) should be assigned the following structure:17

This analysis opens up a new way to handle adverbs such as badly in

the following example :18
(60) He treated Jill badly.
Alexiadou (1997) argues that complement-type adverbs (completely,
easily, badly, well, lovingly, etc.) are generated in the complement do-
main of the verb. Put differently, such adverbs are taken as optional
complements of V. See McConnell-Ginet (1982) forr related discus-
sion. Cinque (1999) proposes that adverb phrases are generated in the
specifier position of their relevant functional heads. Their intuitions
may be reoriented if we extend the analysis of (59) to (60) by assuming
the following structure:
(61) [xp [he treated Jill]i [x'X [AdvPti [Adv'badly]]]]
Taking parallelism strictly, badly is not a complement, but it is actually
predicated of he treated Jill. There is a functional head related to the
adverb. The "matrix" clause is moved into the specifier position of the
functional head.

6. Conclusion
The book under review attempts to resolve the long-standing ques-
tions concerning the categorization of subordinating conjunctions, the
syntactic format of subordinate clause, and the licensing of subordinate
clauses in complex structures. Accounting for these questions,
Haumann offers a novel analysis of "the spectrum ranging from lexical
to functional categories" (p. 147). In this review article, I have ex-
amined Haumann's analysis of subordinating conjunctions, discussing
some consequences and problems. Despite the shortcomings one finds
in it, the book should be praised for its innovative proposals as well as

at LF. Note also that, if P is not involved, the root sentence forms a CP. Thus,
in (i), the complement of the matrix predicate is taken to be, a CP:
(i) Paul said [CPthat he left two hours].
18 See Higginbotham (1985) for the view that such an adjunct is taken as predi-
cated of events.
 CATEGORIZATION AND LICENSING OF SUBORDINATING CONJUNCTIONS 617

its clear, concise, and well-organized discussion. Though left out in

this review for reasons of space, Haumann's thorough review of the
literature (Part I and Part II) is also praised. The book is certain to
show us where the theory has been, and how future research can ad-
vance. The Syntax of Subordination is thus of great value to anyone
delving into the problems addressed in it.

REFERENCES

Abney, Stenen (1987) The English Noun Phrase in Its SententialAspects, Doc-
toral dissertation, MIT.
Alexiadou, Artemis (1997) Adverb Placement: A Case Study in Antisymmetric
Syntax, John Benjamins, Amsterdam/Philadelphia.

197-219.

Bresnan, Joan (1972) Theory of Complementation in English Syntax, Doctoral

dissertation, MIT.

Cinque, Guglielmo (1999) Adverbs and Functional Heads: A Cross-Linguistic

Perspective, Oxford University Press, New York/Oxford.

Linguistic Inquiry 26, 125-137.

Working Papers in Linguistics 8, 128-172.

Hegarty, Michael (1991) Adjunct Extraction and Chain Configurations, Doctor-
al dissertation, MIT.

Huang, James (1982) Logical Relations in Chinese and the Theory of Grammar,
Doctoral dissertation, MIT.

by Manfred Bierwisch and Karl E. Heidolph, 143-173, Mouton, The

Hague.

tion and Output, MIT Press, Cambridge, MA.

Lightfoot, David (1979) Principles of Diachronic Syntax, Cambridge University
618 ENGLISH LINGUISTICS, VOL. 18, NO. 2 (2001)

Press, Cambridge.

144-184.

Linguistics 15, 97-117.

Ormazabal, Javier (1995) The Syntax of Complementation: On the Connection
between Syntactic Structure and Selection, Doctoral dissertation, University
of Connecticut. Storrs.

tion, ed. by Wim de Geest and Yvan Putseys, 1-9, Foris, Dordrecht.

Reports in Formal and Computational Linguistics 2, University of Geneva.

Roberts, Ian (1993) Verbs and Diachronic Syntax, Kluwer, Dordrecht.

quiry 26, 327-340.

Webelhuth, Gert (1992) Principles and Parameters of Syntactic Saturation, Ox-
ford University Press, Oxford.

Faculty of Economics
Yokohama National University
79-3 Tokiwadai, Hodogaya-ku
Yokohama-shi
Kanagawa 240-8501
e-mail: ynaka@ynu.ac.jp