Document generated on 01/08/2022 7:58 p.m.

Revue québécoise de linguistique

θ-Role Assignment in Derived Nominals

Joseph Emonds

Linguistique théorique. Hommage à Judith Mc A’Nulty Article abstract

Volume 15, Number 2, 1986 Chomsky and Stowell attribute an asymmetry in noun and verb complement
systems to how these categories assign Case. Here, Case assignment is
URI: https://id.erudit.org/iderudit/602562ar dependent on structural configurations that θ-role theory determines; it is the
DOI: https://doi.org/10.7202/602562ar latter which explains asymmetries in the noun and verb complement systems.
Several predicate attribute and Vk (clausal) complement types are examined,
See table of contents and all confirm the θ-role theory in which V can assign θ-roles directly to
sisters of any phrasal type, while N and A can assign θ-roles only "indirectly," to
Ymax appearing in PP structures. This predicts the new patterns examined
here, as well as the asymmetry in how objects of verbs and derived nominals
Publisher(s) are syntactically realized.
Université du Québec à Montréal

ISSN
0710-0167 (print)
1705-4591 (digital)

Explore this journal

Cite this article

Emonds, J. (1986). θ-Role Assignment in Derived Nominals. Revue québécoise de
linguistique, 15(2), 91–106. https://doi.org/10.7202/602562ar

Tous droits réservés © Université du Québec à Montréal, 1986 This document is protected by copyright law. Use of the services of Érudit
(including reproduction) is subject to its terms and conditions, which can be
viewed online.
https://apropos.erudit.org/en/users/policy-on-use/

This article is disseminated and preserved by Érudit.

Érudit is a non-profit inter-university consortium of the Université de Montréal,
Université Laval, and the Université du Québec à Montréal. Its mission is to
promote and disseminate research.
https://www.erudit.org/en/
 0 - R O L E ASSIGNMENT
1
IN D E R I V E D NOMINALS
Joseph Emonds

In recent writings in grammatical theory, a property of noun phrases

termed ' 'abstract case" has been accorded a central role in explaining various
syntactic phenomena. In fact, case has been treated as a syntactic primitive, in
that it is taken to be a fundamental property distinguishing nouns and verbs, in
terms of which other properties are to be explained.
In particular, Chomsky (1981, Ch. 3) and Stowell (1981, Ch. 3) at-
tribute an asymmetry in the noun and verb complement systems to how
these two categories assign case. They claim that subcategorization and
9-role assignments for NP complements is parallel in the two systems, with
any irregularities being typical of the kind o f variation found across lexical
entries. However, since nouns do not assign case, a deep structure sister NP
of a noun can receive case (and hence be interpreted) only by virtue o f a
preposition of, which can assign case.
In the view that I will defend here, and have developed in more detail in
Emonds (1985, Chs. 1 and 5), abstract case assignment is dependent on
structural configurations that the theory of 9-role assignment determines,
and it is the latter which explains asymmetries in the noun and verb comple-
ment systems. That is, the theory of 9-role assignment, and not that o f case,
is the primitive system not always parallel to subcategorization; in turn, case
assignment is completely predictable in terms of the structural configura-
tions provided by 9-theory.

1. Most of this material is taken from section 1.7 of Emonds (1985), and is reprinted with
permission. I presented this material at the Université du Québec à Montréal in April, 1984,
during a visit to Montreal for which Judith McA'Nulty was one of my hosts. She later gave me
extensive and valuable comments on this material (and on the rest of Chapter 1 of Emonds,
1985), almost all of which I incorporated in one way or another into the final text. In fact, I am
deeply indebted to Judith for her careful and sympathetic readings of much else of what I have
written.
92 JOSEPH E . EMONDS

It is not my purpose here to examine the "base-dependent" case theory

proposed in Emonds (1985, Ch. 1); rather, I wish to justify my claim that a
principle of 6-role assignment is the proper device for explaining noun/verb
complement system asymmetries.
To begin with a familiar and much-discussed construction, i f a noun is
subcategorized as + NP in the lexicon (e.g., most derived nominals
of transitive verbs are so subcategorized), the only way this NP can receive a
6-role, in my view, is through a PP structure. That is, the subcategorization
can be satisfied only by a PP sister o f the head N in which the P is empty i n
deep structure. This empty P is typically realized as of in surface structure,
although a principle o f derived structure deletes the empty P in certain
2
transformationally altered constructions.

Thus, contrasting with the V, NP sister configurations for verbs in (1), we

find in the N system (2) either a PP structure with an empty or entry-particular
P for direct objects, or that the feature + NP does not carry over
unaltered to the derived nominal.

(1) describe the city; promise reform; answer a question; blame the
accident on someone; marry Sue; expect this bad news; receive
a phone call.

(2) the description o f a city; the promise o f reform; the answer

[to/*of] a question; the blame [for/*of] the accident; John's
marriage [to/*of] Sue; * John's anxious expectation o f this bad
news; *Mary's reception o f a phone call.

For Chomsky and Stowell, the differences between the V's i n (1) and
the N's i n (2) is one o f case-Marking (3); pairs like describe/description and
promise/promise are all + NP i n the lexicon. Both verbs and deri-
ved nominals appear in the deep structures i n (4), and case is assigned in the
derived nominal by (5).

2. The discussion of when elements such as this P may be empty is contained in the first ap-
pendix to Chapter 2 of Emonds (1985). The formal statement of when these elements may be
empty is given there as follows :
Empty Head Principle. If an empty head X ° induced by sub-categorization c-commands an
m a x
adjacent empty and caseless Y , X ° has no phonetic realization.
"Induced" is a technical term defined and motivated in the reference.
 0-ROLE ASSIGNMENT IN D E R I V E D NOMINALS 93

(3) Only V and P can assign case to complement NP's.

X N P
(4) X ( = N , V ) [ ]
(5) 0 > p [of] / N NP

For me, the difference between (1) and (2) is accounted for by the
following principle o f more general empirical coverage, as will be argued i n
3
what follows.

(6) Direct 0-role Assignment : 9-roles can be assigned only to those

phrasal complements which are sisters to V or P.

The data in (l)-(2) is compatible with (3) or (6), so we need to consider

other constructions to see which is more adequate. Consider first the " l i n k -
i n g " verbs which take predicate nominals : be, become, remain, stay, ap-
pear, etc.

(7) John became an adult.

As is well-known, such verbs do not assign morphological accusative case in

case-making languages, and so it is plausible to assume they do not assign
abstract case to the predicate nominal either. Now, how is (7) interpreted?
Typically, the subject position o f linking verbs is one to which NP's are
moved (via passive and subject-raising), and is hence not a position to which
verbs assign 6-roles. In a sentence like (7), we therefore expect that the sub-
ject NP is interpreted not by virtue of receiving a 0-role from become, but
by virtue o f being the subject o f the predicate nominal adult. The predicate

3. For a discussion of problems for Chomsky's Projection Principle posed by (5), see Yim
(1984). In my framework, of is the P which is inserted when the following NP is "genitive"
(i.e., case-marked by a neighboring SPEC(N)). In the generalized case-marking theory
developed in Emonds (1985, section 1.8), SPEC(N) is the only category capable of assigning
case within an NP; thus, the counterpart to the "structure-building" (5) is (i) :
(i) [0] — • of /
P NP g e n i t i v e

"Genitive case" is related to S P E C ( N ) as "accusative" is related to V.

The theory of grammar must exclude a P P realization of the feature + NP whenever
an object NP can be realized as a sister to the head. This is explicit in Emonds (1985, C h . 1),
where Direct 0-role Assignment (6) takes precedence over what is called there Indirect 6-role
Assignment (via an intervening P ) . This is also made explicit in (5) by stipulating the category
N (and not V); but the Chomsky-Stowell theory might also replace this rule-particular stipula-
tion with a more general principle; e.g., Chomsky's suggestion in 1984 class lectures of a "last
resort principle" could apply to rules like (5).
94 JOSEPH E . EMONDS

nominal NP, on the other hand, receives its case from its subject (as is
shown i n an overwhelming number o f morphologically case-marking
languages) and a 0-role from the verb. (It may be more accurate to say that
the subject-predicate nominal pair, o f which the predicate nominal is the
logical head, receives a 0-role from the verb.)
What do we expect when predicate nominal constructions appear i n
derived nominals? Since a linking verb does not assign (accusative) case to
the predicate nominal, any asymmetric case-marking abilities of nouns and
verbs should not come into play inside the derived nominal. I f (3) were suf-
ficiently general, there would be no reason to expect that derived nominals
with nominalized linking verbs should be excluded or should require a
"case-marking" P.
In contrast, i f Direct G-role Assignment (6) is correct, then derived
nominals with predicate nominals should be excluded, or should require the
insertion of a P which licenses 6-role assignment. I n fact, this latter situa-
tion obtains :
(8) *Each being -J of/as [ a good friend was much appreciated.
* I was disappointed by John's unexpected remaining a cook.
*The becoming [of/as| an adult entails responsibilities.
•Her ten-year stay a political prisoner ruined her career.
(Cf. Her ten-year stay behind bars ruined her career.)
*My arrival a poor man surprised my family.
Her appearance jas/*#j» the unwelcome guest was embar-
rassing.
The chair's resemblance | t o / * 0 } a couch is surprising.

The accusative case-marking asymmetry o f N and V in the Chomsky-

Stowell system does not explain why derived nominals never tolerate
predicate nouns. But the requirement that complements within N receive a
6-role only by being within a PP does.
Predicate adjectives are not case-marked by V either, as a survey of
languages which mark case morphologically also shows. But again we find,
asymmetrically, that A P complements to V are tolerated, while those to N
are not. This again suggests that the ability to case-mark is not the source o f
4
discrepancies in the V and N complement systems.
4. Superficially complicating the issue here is the existence of many derived nominals which
paraphrase verb-predicate attribute combinations :
 0-ROLE ASSIGNMENT IN D E R I V E D NOMINALS 95

(9) John appeared reluctant to leave.

*We were surprised by John's appearance reluctant to leave.
That dessert tasted sweeter than candy.
T h a t dessert's taste sweeter than candy overwhelmed us.
His plant grew tall.
*His plant's tall growth is easily explainable.
M y friend stayed sober for years.
* I am happy about my friend's stay sober for years.

Further confirming evidence for my view that N and V assign 0-roles

differently (and that asymmetric case marking is only a special case o f the
more general contrast) can be obtained by examining predicate attributes
(NP's and AP's) o f transitive as well as intransitive verbs. I n (10), the head
of the V has two phrasal sisters which receive a 0-role directly, one from the
head (V) and the other from the complement o f which it is the subject N P .
The direction of direct 0-role assignment is diagrammed in (11).

a. She could recognize that the trumpet sounded flat.

b. She could recognize the trumpet's flat sound.
But in cases like (b), the trumpet's flat sound is also a paraphrase of " N — is — A " (the
trumpet's sound was flat), and some interpretive device I D must account for this latter alterna-
tion independantly of "V-predicate attribute" combinations, as shown in (c)-(d) :
c. She could pick out the trumpet's flat note.
The trumpet's note was flat.
•The trumpet noted flat.
d. We disliked that flat rendering of the Davis tune.
That rendering of the Davis tune was flat.
•They rendered the Davis tune flat.
Not only can the device I D work on N + A when the corresponding V is not + A P , it
may not work on N + A when V , + A P exists but N-Zs-A does not :
e. * John's appearance was reluctant to leave.
•His plant's growth was tall.
It must be concluded then that the interpretation of A + N combinations as in (a-b) does not
result from any parallelism with V + A P combinations, and hence does not involve a similar
assignment of 6-roles. That is, the adjectives that modify nouns are not arguments of the
nouns, but rather modifiers of a different sort. Both in traditional and early transformational
grammar, it was assumed that these adjectives are directly related to relative clauses, at least in
their mode of interpretation; as such, they do not receive 0-roles from N, but are interpreted by
some independent mechanism.
I retain this assumption here, as there is no evidence to contradict it. With this assumption
providing the explanation for the existence of apparent counterexamples like (a-b), the ex-
amples in (9) and (12) provide confirming evidence that N does not directly 0-mark A P , while
V does.
96 JOSEPH E . EMONDS

(10) We elected John secretary.

Chomsky considered that paradigm an interesting problem.
I judged this the best entry.
They appointed me guardian o f your estate.
They made me an illegal alien.
Who classified them political prisoners?
He considers that law repressive.
She dislikes reggae music loud.
Bill might prefer his steak rare.
Kathy proved him wrong.
Many believed a Labor Party capable of redistributing wealth.
They recommended preparing the meat dry.

(11)

consider that law repressive

.0,

Since complements to N can be assigned 0-roles only via an i n -

termediate P, it follows from (6) that derived nominals exactly correspond-
ing to the examples in (10) should be excluded. They are so excluded :

(12) *Our election of John secretary was illegal.

•Chomsky's consideration o f that paradigm an interesting
problem was a turning point.
*My judgment o f this the best entry was criticized.
T h e i r appointment o f me guardian o f your estate was a
mistake.
T h e i r making of me an illegal alien was unprecedented.
*Any classification o f them political prisoners would be a step
forward.
 0-ROLE ASSIGNMENT IN D E R I V E D NOMINALS 97

*The consideration o f that law repressive is evidence of an

open mind.
*Her dislike o f reggae music loud cost her a friend.
•Bill's preference o f his steak rare came as no surprise.
*Kathy's proof o f him wrong came at a good moment.
*A belief o f a Labor Party capable o f redistributing wealth was
a post-war characteristic.
T h e y recommended the preparation o f the meat dry.

The theory of 9-role assignment does not determine of itself when the use of
the prepositions of and as together suffices to create a derived nominal cor-
5
responding to an example of (10). Direct 0-role Assignment only predicts
that i f such a derived nominal exists, all the 0-role assignments within it will
be indirect. So there are acceptable examples as in (13) :

(13) Our election o f John as secretary was illegal.

Your opinion o f that law as repressive is evidence o f an open
mind.

In Emonds (1984), I show that non-comparative as, as i n (13), is indeed

a preposition, but one whose peculiarity, like that o f the verb be, is that it
does not assign case to its sister NP. Such a P (like the verb be) does,
however, allow its sister to be assgined a 0-role, and hence to satisfy (6).
When no alternative with as is available, the derived nominal for a verb with
a subcategorization feature + NP A P or + NP NP either
6
does not exist, or is not compatible with this complement structure.

5. Direct 6-role Assignment does not itself specify that a P rather than a V or some subor-
dinate clause strategy be used to express a complement to an N . As suggested by Ian Roberts
(pers. comm.), a language with serial verbs such as Chinese might be one which chooses V
rather than P as a category to introduce complements when direct 0-role assignment is not
available.
6. Chomsky (1970) briefly discusses "action nominalizations," which for the most part ex-
hibit internal structure typical of NP's, and whose head is of the form W-ing. Predicate at-
tributes in such nominals are marginally acceptable :
? Their painting of the White House red disturbed even the Secretary of Labor.
?His calling of the rebels Communists gave the signal to the death squads.
The restrictions that action nominalizations are subject to, such as the above and several others
pointed out by Chomsky, suggest to me that they are derivatively generated and hence ungram-
matical, for reasons entirely analogous to those given by Chomsky (1970) for derivatively
generating noun-modifying adverbial clauses.
98 JOSEPH E . EMONDS

The reader may verify that derived adjectives corresponding to verbs

which take predicate attributes do not appear with A P and NP com-
plements either. Just a few typical examples are given here.

(14) The desserts are tasty. The desserts taste too sweet.
*He criticizes desserts tasty too sweet.
We are considerate o f our guests.
We consider our guests very comfortable.
*We were considerate of our guests very comfortable.
Mary was judgmental. Mary judged John ill-tempered.
*Mary was judgmental o f John ill-tempered.

So far, I have shown that N's and A's cannot assign 0-roles directly to
NP and A P sisters, whether they are direct objects or predicate attributes.
I n Emonds (1985, C h . l ) , I go on to explain how the asymmetric case-
marking properties o f N and V follow automatically from the asymmetries
7
imposed on deep structures by G-role theory. Here, I will continue
demonstrating that an elaboration o f case-marking theory fails in principle
to explain further asymmetries of complement structure between V on the
one hand, and N and A on the other.
In particular, I will investigate whether N can assign G-roles directly to a
projection o f V . Direct G-role Assignment (6) again predicts an asymmetry
in the array o f possible clausal complements to N and V . V and P should be
k
able to take V clausal complements, where k = 2 or 3, without benefit o f
an introductory grammatical formative o f the category P, but N should not
k 8
be able to assign a G-role directly to a V sister.

The first such gap i n complements to N's that concerns us has to do

with a restricted kind o f non-finite complement in English introduced by V

7. One such a symmetry is that in a case language like German, the NP objects of a transitive
verb are typically accusative, while prepositionless NP complements to nouns and adjectives
are invariably marked as dative or genitive rather than accusative (van Riemsdijk, 1983). As
argued in Emonds (1985, C h . 5), this oblique case inflection within NP's is best explained by
the presence of an empty deep structure P (as also with indirect objects of a verb). Thus, as far
as phrasal sisters to X ° are concerned, all and only the NP which are sisters to X ° are in fact
sisters of V or P.
2
8. In Emonds (1985, C h . 3), I present arguments that a V P is best analyzed as V and that S
3 2
is best analyzed as V , in a system where other maximal projections such as NP are N . For
purposes of the argument here, however, it only matters that the various types of clausal com-
plements discussed in the text are not NP's or A P ' s .
 6-ROLE ASSIGNMENT IN D E R I V E D NOMINALS 99

-I- ing. I n a previous study o f English gerunds (Emonds, 1976, Ch. IV), I
gave a series o f syntactic tests in support o f a hypothesis o f Rosenbaum
(1967) that gerundive complements are NP's, with (only) two classes of ex-
ceptions : the V + ing complements o f intransitive verbs o f temporal aspect
are not NP's (examples italicized in (15)), because they do not undergo
various NP movements and otherwise do not behave as NP's. For similar
reasons, both the V + ing complements and certain bare infinitives after
transitive verbs o f perception (italicized in (16)) are not NP's.

(15) They continued clearing the street.

Did David start doing his project?
She should cease describing the machines.
(16) One should see a cat fight(ing) another cat.
They noticed me tak(ing) a tooth brush off the rack.
They will arrest Bill picketing the consulate.

My claim is not that a cat fighting another cat can't be an NP gerund,

but rather that, under one structural analysis, this string is not an NP
(following Chomsky (1957); cf. also Akmajian (1977)). Two arguments for
this are : (i) The NP understood as the subject of the perception verb in-
finitive or gerund (e.g., a cat) can be extracted, in contrast to the subject o f
an NP gerund, which cannot be :

(17) What did they see fight their cat?

Who did they notice taking tooth brushes?
•What did they describe fighting their cat?
•Who do they enjoy playing the piano?

(ii) Essentially any NP can be the focus constituent in a cleft sentence, but
the VP following a perception verb cannot be so focused :

+
(18) I t was picketing the consulate that they arrested Bill.
•It's taking tooth brushes that they notice shoplifters.

In Emonds (1985, Ch. 2), I argue that the gerund complements of in-
transitive temporal aspect verbs and o f transitive perception verbs are VP's,
not S's, in deep structure. I n the context of this argument, whether they are
VP or S complements is not crucial; as long as they are not NP's or AP's,
100 JOSEPH E . EMONDS

they do not receive abstract case. Hence, a difference in their distribution

within V and N cannot be attributed to differing case-marking properties o f
the head V and N .

k
Direct 0-role Assignment (6) predicts that the V complements o f tem-
poral aspect and perception verbs (k = 2 or 3; see note 8) cannot be inter-
preted as sisters to N , and this prediction is borne out :

(19) *The continuation (of) clearing the street was a surprise.

*We were all relieved at David's start (of) doing his project.
*A cessation (of) describing the machines would be welcome.
*The sight o f a cat fight another cat is interesting.
•Their notice of me take tooth brushes led to my arrest.
•The arrest of Bill picketing that house went unnoticed.

French lacks gerund complements (verbal forms i n -ant appear only in

adverbial positions), but it does seem to have infinitives that correspond to
the non-finite constructions in (15)-(16). French infinitives can follow tran-
sitive perception verbs; however, the counterparts to the temporal aspect
complements are possibly exemplified by the complements to a different
semantic class o f verbs. The initial V in these structures are the motion verbs
such as monter, descendre, sortir, partir, etc. I t has been noted i n the
literature (e.g., Gross (1968)) that the prepositionless infinitives that follow
these verbs may not be negated, passivized, or otherwise modified.

(20) Michel est sorti acheter du v i n .

'Michael went out to buy wine.'
Marie va descendre voir ses amies.
'Mary will go down to see her friends.'
Elle part faire du tourisme.
'She's leaving for some touring.'

2
By analyzing these infinitives as VP (V ) complements to V, the basis is laid
for an account o f their inability to passivize, to be negated, etc. In any case,
these infinitives do not have the properties of NP's.

The evidence for (6) consists in the fact that the derived nominals o f
these motion verbs cannot be followed by such infinitives :
 6-ROLE ASSIGNMENT IN DERIVED NOMINALS 101

(21) Sa sortie (*acheter du vin) n'a pas été remarquée.

'His leaving (to buy wine) was not noticed.'
Sa descente (*voir ses amies) sera probablement périlleuse.
'Her going down (to see her friends) will probably be
dangerous.'
Son mari est triste de son départ (*faire du tourisme).
'Her husband is sad over her leaving (for some touring).'
The impossibility o f direct 0-role assignment between a derived nominal and
a projection o f V (either VP or S) explains the contrast (20)-(21).

Both French and English have infinitival complements which are pro-
3
bably best analyzed as V (S) complements. French infinitives and English
infinitives with lexically realized subjects (impossible in French) both pro-
vide evidence i n favor o f the contention that P and V assign 0-roles directly,
while N and A assign them indirectly.
Huot (1981) argues that the morphemes à and de which introduce
French infinitives are members of COMP. In Emonds (1985, Ch. 7), I argue
that COMP are P, so as a special case the French à and de before infinitives
are then P also; this is o f course corroborated by the fact that à and de are
also P's which appear in the context + NP in French. (6) then
predicts that N and A can only have infinitival complements introduced by
a P (such as à and de), while some V may be subcategorized to take in-
finitives without a COMP. Again, this prediction o f (6) is straightforwardly
borne out :
(22) I l préfère (*à) boire du vin blanc avec le poisson.
'He prefers to drink white wine with fish.'
Sa préférence j j boire du vin blanc avec le poisson a
été encouragée.
'His preference for white wine with fish was encouraged.'
Claire a voulu (*de) changer de travail.
'Claire wanted to change her j o b . '
La volonté de Claire j de/*^ ^ changer de travail n'a
pas été respectée.
'Claire's wish to change jobs wasn't respected.'
I l dit pouvoir influencer ses parents.
'He says he can influence his parents.'
On a discuté son pouvoir | d ' / * ^ j influencer ses parents.
'They discussed his ability to influence his parents.'
102 JOSEPH E . EMONDS

Since S infinitives do not receive case, the contrasts in (22) can not be
predicted by the inability of N to assign case.
Turning again to English, the subject NP of an infinitive may be lexical-
ly realized after a relatively small set o f verbs o f belief and desire. For a
variety of reasons, Chomsky (1981, Ch. 2) argues that i n such structures, no
S can intervene between the verb and its S complement. Examples of such S
9
complements are italicized in (23).

(23) He expected there to be lasting peace.

We prefer the weather to be cool.
Anne considers her travels to be indispensable.

The impossibility of direct 0-role assignment by N in the present system

implies that derived nominals corresponding to (23) can contain clausal
complements only i f they are introduced by an intervening COMP (i.e., a
P). A n d this is the case :

(24) *His expectation (of) there to be a lasting peace was never met.
*Our preference (of) the weather to be cool should be taken
into account.
Our preference for the weather to be cool should be taken into
account.
* I am aware of Anne's consideration (of) her travels to be in-
dispensable.

9. Chomsky refers to these complements as being derived by "S-deletion", indicating that

at deep structure, there is perhaps an S present. I believe that the only motivation for an S
might be to insure the possibility of long-distance WH-movement out of such complements via
a C O M P "escape hatch." If there is no S present, and if S is a bounding node for subjacency,
multiple levels of the type of complement in (23) could lead to subjacency violations, and yet
extractions in such configurations are acceptable :
(i) How warm did he expect us to want our room to be?
Granting both subjacency and that S is a bounding node in English, the fact still remains that
extractions out of two adjacent bounding domains of the same category type produce examples
that are either acceptable or only very slightly unacceptable. Thus, (i) and (ii) have the same ac-
ceptability, as far as I can tell.
(ii) Which city did you take pictures of the outskirts of?
We went along a river that the trees along the banks of looked very unhealthy.

Thus, I see no reason to assume that the complements of the main verbs in (23) are S's at any
level of description.
 0-ROLE ASSIGNMENT IN D E R I V E D NOMINALS 103

Having now seen instances o f both VP and S non-finite complements to

V whose counterparts are lacking inside N , we can turn to whether there are
finite complements with the same skewed distribution, as principle (6)
predicts. I n most recent generative work, any embedded S is assumed to be
generated by the following rule.

(25) S • C O M P —S

As mentioned earlier, I argue elsewhere that S should be identified with PP

and that the grammatical formative category C O M P should be identified
with P. Given this result, the above question becomes, can V take a finite
clause directly, without an intervening COMP and S (just as P = COMP
can), while the categories that cannot assign a 0-role directly, N and A , are
obliged to have only S complements?
As seen above, the answer to the corresponding question for a range o f
different types o f non-finite clauses is yes. For English, the same answer is
warranted for finite clauses, since the COMP can be absent between certain
V and a finite complement :

(26) John feared Mary would be late.

•John's fear Mary would be late turned out to be unjustified.
She decided no one qualified.
•Have you heard about her decision no one qualified?

In my terms, the C O M P ( = P) that is providing [ P - S ] structure, which is

PP

allowed for both V and N ; however, V can also have an S sister which is
0-marked directly. Thus, (6) correctly predicts that that will appear op-
10
tionally with verbs, but that it must appear with nouns.
10. In French, there is no large class of verbs where que 'that' can be omitted before the
complement S. But significantly, with only one or two exceptions involving W H (si ' i f , quand
'when'), que is also always required between subordinating conjunctions P and a complement
S : pendant que 'while', avant que 'before', puisque 'since', lorsque 'when', bien que
'although', dès que 'as soon as', etc. If we make the plausible assumption that que is inserted in
French with every finite S, the following rule can be postulated, essentially as part of the map-
ping from surface structure to phonological form :
0 » que I C finite S; C £ W H ; obligatory
If C = N or C = A , there is an empty P required by Direct 0-role Assignment between N , A
and a finite S, so que will fill that P in phonological form. If C = a lexical P, there will be no
such empty P following, and if C = V, there need not be. In these latter cases, que will be in-
serted into the terminal string but will not be assigned a syntactic category. With this clarifica-
tion for French, it appears that the prediction that V and P assign 0-roles to S complements in
parallel fashion can stand (both for French and English).
104 JOSEPH E . EMONDS

k
In this study, several instances of predicate attribute and V (clausal)
complement types have been examined, and shown to bear our the theory o f
0-role assignment expressed in (6). I n particular, these constructions taken
together demonstrate that V can assign 0-roles directly to sisters of any
phrasal type, while N and A can assign 0-roles to complements only in-
m a x
directly, to Y which are PP or are embedded i n PP. Since predicate at-
tributes, whether NP's or AP's, are not assigned case by either a governing
verb or noun, case theory cannot predict the patterns observed here, while
0-theory, as expressed in (6), predicts not only the new patterns examined
here, but also the long-noticed asymmetry in how objects to verbs and
derived nominals are syntactically realized.

The clausal complement structures which support the theory of 0-role

2
assignment presented here are : (i) the V participial clauses which appear
after verbs o f temporal aspect but not in derived nominals (English); (ii) the
2
V participial and infinitival constructions which appear with transitive
perception verbs (English and French), but not after the nominals derived
3
from these verbs; (iii) the French subjectless V infinitives which can be i m -
mediately preceded by a governing V or P, but never by a governing N or A
(an intervening [à] or [de] being necessary); (iv) the English infinitives
p p

with lexical subjects after verbs like expect and prefer, excluded after cor-
3
responding derived nominals; and (v) the COMP-less finite V complements
which are sisters to verbs like fear and decide (English). Since, in the case-
k
marking system devised by Stowell (1981, Ch. 3), the various V never
receive case, the asymmetry i n their distribution cannot be properly con-
sidered as due to the theory of case-marking. Rather, the more general prin-
ciple of Direct 0-role Assignment (6) developed here is the appropriate ex-
planation for a much wider range o f facts. Differences i n case-marking
follow from the influence of 0-role theory, rather than being stipulated as
primitives in Universal Grammar.

I conclude that every complement inside N or A must be introduced in-

side a structural PP. This result dovetails with the independent arguments
given in Emonds (1985, Ch. 7) to the effect that C O M P is a P and S is a PP,
since we have seen that clausal complements to N often require a COMP
while those to V do not ((iii)-(v) above).

From these considerations, there emerge the following two points, both
essentially implicit i n traditional grammar.
 8-ROLE ASSIGNMENT IN D E R I V E D NOMINALS 105

(27) V and P are "relational" categories; each can introduce any

maximal phrase as a complement (the maximal phrases are NP,
A P , VP, S, and PP). N and A are non-relational; their com-
plements must be introduced via the universal subordinating
category P.
(28) Abstract and morphological case principles are not asymmetry-
inducing primitives in Universal Grammar, but are indicators
of the presence of local (usually adjacent and sometimes emp-
ty) case-assigning categories. Case assignment is based on the
following pattern :
accusative : presence of V
dative/oblique presence of P
genitive : presence of SPEC(N)
!
nominative : presence of SPEC(V) = " I N F L

In (28), the informal term "presence o f " is in fact to be replaced by the term
"governed b y " , where the appropriate definition of government is as in
Emonds (1985, section 1.8).

Joseph Emonds
University of Washington
Stanford Center for Advanced
Study
106 JOSEPH E . EMONDS

Références

A K M A J I A N , A . (1977) «The complement structure of perception verbs in an autonomous

syntactic framework», in P. Culicover, T . Wasow, and A . Akmajian (eds) Formal syn-
tax. Academic Press, pp. 427-460.
C H O M S K Y , N. (1957) Syntactic structures. Mouton.
C H O M S K Y , N. (1970) «Remarks on nominalization,» in R. Jacobs and P. Rosenbaum (eds)
Readings in English transformational grammar. Ginn, pp. 184-221.
C H O M S K Y , N. (1981) Lectures on government and binding. Foris Publications.
E M O N D S , J . (1976) A transformational approach to English syntax. Academic Press.
E M O N D S , J . (1984) «The prepositional copula as.», in Linguistic Analysis 13, pp. 127-144.
E M O N D S , J . (1985) A unified theory of syntactic categories. Foris Publications.
G R O S S , M . (1968) Grammaire transformationnelle du français : syntaxe du verbe. Larousse.
H U O T , H . (1981) Constructions infinitives du français. Librairie Droz.
R I E M S D I J K , H . van (1983) «The case of German adjectives», in F . Heny and B. Richards
(eds) Linguistic categories; auxiliaries and related puzzles. Reidel.
R O S E N B A U M , P. (1967) The grammar of English predicate complement constructions.
M . I . T . Press.
S T O W E L L , T . (1981) Origins of phrase structure. M . I . T . doctoral dissertation.
Y I M , Y . - J . (1984) Cas-tropism : the nature of phrasal and clausal case. University of Wash-
ington doctoral dissertation.