Ch.9 Aspect and Arg Str2 - Hale and Keyser
Ch.9 Aspect and Arg Str2 - Hale and Keyser
Ch.9 Aspect and Arg Str2 - Hale and Keyser
And this structural difference accounts for their behavior in relation to the
standard causative-inchoative transitivity alternation:
(3)
(4)
The properties which distinguish these two verbs are the following. The verb
break, as illustrated in (1) and (3) consists of the following structural elements, a
root (R) and a verbal host (V):
(5)
R, V
The verbal component has the property that it takes a complement, realized here
as the root. The latter contains the semantic and phonological features associated
with the dictionary entry break. The root component requires a specifier, as
shown in (6):
(6)
V
DP
the pot
V
R
V {break}
This is an essential feature of the root {R, break}, accounting for the central
syntactic feature of the verb, namely the transitivity alternation observed.
The verb cough, represented in the grammarical sentence (2) and in the
ungrammatical sentence (3), likewise consists of two parts, a root, and a verbal
nucleus. Unlike the root component of break, however, the root element of cough
does not require a specifier, thus the verb does not, and cannot, project a
specifier:
(7)
V
R
V cough
A verb, in and of itself, does not project a specifier, and its complement in this
case (i.e., root element) does not motivate the projection of a specifier. These
properties account for the ill-formedness of (4).
Transitivization of the type represented by (3) is in principle automatic, by
virtue of the complement relation. The structure of (3) is a result of the
combination, via Merge, of (6) and a verbal nucleus V, as in (8):
(8)
V1
V1
V2
DP
the pot
V2
R
V2 break
V2
R
V2 cough
This follows from the fundamental nature of the root {cough}, which does not
force the verb to project a specifier. In general, but with some exceptions, this
2
(a)
(b)
(a)
(b)
P
DP
a quarter horse
P
N
P saddle
DP
make a fuss
(a)
(b)
(c)
The first of these has to do with the principles involved in the circumstance that
the phonological matrix associated with the nominal root cough is realized in the
verb of (2), and not in its complement. The second problem is semantic in nature.
It has to do, among other things, with the fact that the semantic features of the
root component of a verb are sometimes linked with an internal argument (object
or specifier) and sometimes with the external argument (the sentential syntactic
subject). The consequences are straightforward in the syntactic behavior of the
relevant verbs. The third problem involves an issue with which we have not
dealt heretofore, although we have alluded several times to an opposition (i.e.,
central and terminal coincidence) which may be relevant. The problem will be to
determine the role of structure in this domain.
1. Conflation and Selection.
Conflation is a term that we use to refer to the phonological instantiation
of light verbs in denominal verb constructions. Specifically, the issue of
conflation has to do with the problem of how the verb ends up carrying the
phonological matrix of its nominal complement, as in examples of the type
represented by (2) and (10a) above, the relevant structures for which are repeated
here:
(15)
(a)
V
R
V cough
(b)
V
V
P
DP
a quarter horse
P
N
P saddle
These representations give the impression that the basic structures locate the
phonological matrix of the noun in the noun itself, i.e., in the complement of V in
the case of (15a), of P in the case of (15b). On this view of the matter, which we
held to be self-evident for many years, the spell-out of the verb (cough, saddle, in
these examples) required a kind of movement, resulting ultimately in the
acquisition by the V of the phonological matrix of the relevant noun. It seemed
reasonable to propose that the movement operation involved in these derivations
was Incorporation, in the technical sense of Baker (1988). This idea was
abandoned, however, because incorporation overgenerates, incorrectly
sanctioning incorporation from the position of the internal specifier (e.g., from
the position of DP in (15b)). Unconstrained, incorporation will permit forms like
those in (16):
(16)
(a)
(b)
(17)
(a)
(b)
P
N
salt
P
P N
box
V
V
cough R
box
N
salt
P
P N
(a)
(b)
He danced.
He danced a jig.
6
Verbs which are candidates for Conflation appear in constructions like (20a); that
is why they are candidates for Conflation. But virtually all such verbs also appear
in constructions like (20b), in which the phonological matrix (dance in this case)
must be considered in some sense "basic." While dance could be derived from the
complement in (20a), it is not obvious how it could be derived from the
complement in (20b). In short, we must assume that the verb dance is entered as
such in the lexicon, complete with its full phonological matrix.
This challenges the basic foundations of Conflation as a theory of
phonological realization. Except as an item of terminology, Conflation ceases to
exist. The relation subsumed by this term reduces to another fully established
and generally recognized relation, namely Selection. We have already suggested
that Selection is a condition on Conflation. Suppose we carry this thought further
and simply identify Conflation with Selection, folding the former into the latter.
The idea would be that the full verb of (20a), for example, would be "rich
enough" in semantic features to license the empty category functioning as its
complement. This is a kind of selection, inasmuch as the verb identifies the
empty category as a hyponym of "dance," i.e., a member of the class of entities
which qualify as dances. This conception of the matter has the advantage that the
semantic relation involved in (20a), where the complement is non-overt, is
essentially the same as in (20b), where the complement is overt. The overt
complement, a jig, is identified as a hyponym of "dance," i.e., the jig which is a
dance, as opposed to a musical score, a fiddle tune, or whatever else "a jig" might
mean.
There is some support for this from the licensing of non-overt
complements. A non-overt complement is possible if it is selected in the sense
indicated, i.e., identified as a hyponym by semantic features inherent in the
governing verb. It follows then that so-called "light verbs" cannot license a nonovert complement:
(21)
(a)
(b)
(c)
(a)
(b)
(a)
(b)
V
V
V
P
splash
DP
mud
P
P DP
DP
mud
V
splash
P
P DP
o n the wall
o n the wall
The two alternants (Hale and Keyser, 2000b) are defined straightforwardly and
automatically by the operation Merge (Chomsky, 1995). Ceteris paribus, the
8
(a)
(b)
(25)
(a)
*V
P
smear
DP
mud
DP
mud
P
P DP
P
P DP
smear
o n the wall
o n the wall
The difference between these two verbs lies in the semantic components of
their root elements. Specifically, the difference is to be found in what might be
termed the "manner factor" inherent in the semantics of the root. The verb splash
in (22) involves a manner feature which is in a clear sense "linked" to the internal
argument mud. It represents the motion and dispersal of particulate matter
associated with mud, not with the external argument. This relation is preserved
in both the transitive and the intransitive alternants. By contrast, the verb smear
in (24) is characterized by a "manner feature" linked externally, i.e., embodying a
gesture or motion associated with the external argument. This relation is, of
course, disrupted in the intransitive alternant depicted in (25b). The smear factor
cannot be linked to the external argument there, since that position will be taken
by the internal argument, raised there in sentential syntax. Accordingly, the
sentence is ungrammatical (Hale and Keyser, 1999).
Verbs of impact and concussion behave in a related manner; the following
is a relevant pair:
(26)
(a)
(b)
These are assumed to have the locatum structure (cf. give it a kick):
(27)
(a)
(b)
V
V
V
P
DP
the wall
P
N
P kick
P
DP
the fender
P
N
P dent
(a)
(b)
(a)
(b)
(a)
(b)
10
V
DP
V
A
V clear
11
As usual, this diagram represents the properties of the heads involved. It is the
virtual structure, not the actual outputMerge applied to V and A results
immediately in Conflation, giving the verb clear, as in the sky cleared.1
But what of the adjective when it appears to lack a host for the specifier it
requires? Consider, for example, the structure of an adjectival small clause, of the
type illustrated in (32):
(32)
(a)
(b)
(c)
(d)
If the sky in (32a) is in a specifier position, what head projects that position? We
have assumed that A itself does not merge directly with the phrase that satisfies
its specifier requirement, since the resulting relation would be indistinguishable
from that holding between a head and its complement, not the required relation
here. And in (32) there is no other obvious candidate to host the specifiera
problem, on the face of it. The solution can be seen by considering the difference
between conflation constructions like (31) and free-standing adjectival predicates
like those in (32).
In the conflation construction, the adjectival component is an unprojected
headthat is to say, a bare adjective. In the small clause construction, however,
we assume that the free standing adjective is the lexical head of an extended
projection. In (32), it happens that no part of the extended projection is overt,
since the adjective is in the absolute degree. In the examples in (33), however,
elements of the extended projection are overt:
(33)
(a)
(b)
(c)
1There
are two kinds of incorporation, in our view. The type which we have termet
tconflation here is strictly a matter of the PF representation of syntactic objects. It is a part of the
operation Merge. When a dependent head (i.e., a phonologically non-overt head or an affix)
merges with a complement, the latter fuses with the former to eliminate the empty phonological
matrix. This fusion, of course, involves the head of the complement, not, say, a specifier of the
complement, accounting for the impossibility of we appled in boxes, beside well-formed we boxed
apples. By contrast, sentential syntactic incorporation generally involves adjunction of a bare root
(noun, verb, etc.) to an overt locally c-commanding verb, resulting in a composition of overt
elements, i.e., a compound. Constraints of incorporation are the familiar ones, including the ECP.
In principal, and in fact, incorporation of a specifier into a c-commanding verb is possible.
Incorporation is also morphologically driven, possibly, since in many languages that have
incorporation (e.g., Tanoan, Nahuatl, Kunwinyku) the bare root cannot stand alone.
12
(34)
DP
the sky
clear
Among the elements which occur in the head position are , the non-overt
head of the absolute degree, exemplified in (32a), and -er, the affixal head of the
comparative degree, exemplified by (33b)these both implicate conflation,
eliminating the empty phonological matrices. Other members of the category
presumably include so, as, too, very. The -projection exemplified in (34) appears
as the complement of a verb in (32a,b) and as the complement of a preposition in
(32c,d). In (35) it appears as the complement of raising predicates, including the
copula:
(35)
(a)
(b)
(c)
The adjective conflates with the phonologically empty head in (35a) and, in (35c),
with the empty matrix associated with the comparative degree suffix -er.
The structures (31) and (34) share the property that they are dyadicin
both cases, the head projects two argument positions, corresponding to the
relations termed complement and specifier.2 There is an important difference
between the two structures, a difference which resides in the nature of the head.
While V and both select adjectival complements and DP specifiers of the same
general sort (appropriate to the adjective), they differ consistently in stativity.
The V-based structure is active (non-stative) and the -based structure is stative.
In this discussion, we will be concerned in large part with the question of
stativity, and with its source and proper representation in the grammar. We
will take a number detours, however, in order to discuss structural matters
which come up. We begin with a consideration of the possibility that stativity
correlates with lexical category or part of speech.
3.1. Stativity and category.
It is not unreasonable to ask whether it is a general principle that verbs
project structures associated with an active (nonstative) interpretation while
other categories project structures associated with a stative interpretation. In
some languages, this is true without exceptione.g., it is true in Warlpiri of
Central Australia. But it is of course well known that, in a great many other
languages, including English, there are verbs which are stative according to
2Since
these are structural relations, the terms complement and specifier have no special status,
being simply the names of the structural relations: (i) a complement is the sister of the head, and
(ii) a specifier is the sister to the syntactic object consisting of the head and its sister. We will
continue to use these traditional terms, nonetheless, as an expository convenience.
13
What accounts for the stativity of these verbs? One possibility is that they
involve the dyadic sturcture projected by the category Pspecifically, the covert
P of central coincidencelike that found in locatum verbs of the type
represented by saddle, hobble, clothe. Accordingly, these verbs would have
paraphrases involving give, as in John gave the truth his respect, or, more accurately
John got the truth (to be) with his respect, where with corresponds to the overt
possessive preposition, a prototypical preposition of central coincidence, also
illustrated in secondary predicates likewith gifts, as inthey came with gifts. Of
course, the preposition putatively implicated in (36) is empty, non-overt, and
necessarily conflates with its complement. Under these assumptions, the dyadic
structure underlying the verb phrases of (36) is as follows (using respect the truth
to illustrate):
(37)
P
DP
the truth
P
N
P respect
14
(38)
(a)
(b)
(c)
(d)
(e)
The structural correlation is this, taking (38a) as the model and comparing this to
(37). The subject of the have-construction, Mary in this instance, corresponds to
DP in (37), and the object of have, i.e., my respect, corresponds to N, complement
of P; have itself corresponds to P. In essence then, the predicates in (38) are
structurally identical to (37). The differences between them are matters of
realization and selection(37) is headed by empty P, whose complement is a
bare N, while the predicates of (38) are headed by an overt, morphologically
verbal element have, whose complement is a full DP, specifically a possessive
construction linked to the external subject.
We will resume this structural comparison at a later point. For the present,
let us return to the issue of stativity. We ask whether the suggested categorial
affiliation of the head of (37) could be the source of the stativity of the verb
phrases of (36). This would be in line with the proposal that non-verbs head
stative projections.
The usual fate of a P-headed structures like (37) is to enter into
construction with another category, as when it appears as the complement of the
lexically monadic V-head, as shown in (39):
V
(39)
V
P
DP
the truth
P
N
P respect
(a)
(b)
(c)
15
assumption that these verbs can in fact enter into the construction presented in
(39), essentially the structure of locatum verbs.
If the stative predicates of the have-constructions of (38) are structural
paraphrases of (37), then the give-construction predicates seen in the slightly
stilted (41) are structural paraphrases of of (39):
(41)
(a)
(b)
(c)
Here again, the difference is one of realization and selectionthe head is overt in
(41), non-overt in (39), and the complement in (41) is a possessive DP linked to
the external argument.
However, if (39) accounts for the non-stative use of experiencer-subject
psych-verbs, what accounts for their allegedly more fundamental stative use, as
in (36)? On the view that the stative counterparts are lexically non-verbal, there is
a rather natural suggestion that can be made. The head of (37), as given, is a nonverbal headits head is P, by hypothesis. By contrast, the head of (39) is verbal.
Of course, the two are homophonous, taking the form respect. But this follows
from the fact that both result from "conflation" of the same bare nominal, respect.
This gives overt phonological form to P, yielding the P-based predicator respect.
The same nominal root gives phonologival form to V in (39), deriving the verbal
variant of respect exemplified by (40a).
If the distinction between stative and non-stative experiencer-subject
predicators like respect, love, fear, etc., can be attributed to lexical category (V, P,
etc.), then the suggestion we are entertaining now could in principle be the
solution to the problem of stativitystatives are P-based, non-statives are
V-based. There is another part of the problem, however. The stative is just as
much a verb, in the traditional sense, as the non-stative is. That is to say,
contrary to what is expressed in (37), the stative variant of respect assumes the
same commanding position that its V-based active homophone does. And like
the latter, the stative variant enters into the same inflectional relations (e.g., tense
inflections) as the non-stative, unquestionably verbal, variant does.
One possibility which might be considered is that the P of (37), while not
itself verbal, must inflect with verbal morphologyin violation, to be sure, of the
principles which generally hold in extended projections (Grimshaw, 1991). If this
morphological eccentricity were in fact a property of P in (37), then its
satisfaction would require P (with conflated N) to be raised to a position from
which it c-commands its original position and those of its arguments. For the
present, let us suppose that P raises and merges with its own maximal projection,
as shown informally in (42):
16
P*
(42)
P*
P^N
respect
P
DP
the truth
P
tP tN
(a)
(b)
(c)
(d)
Thus, the motivation for the putative P-raising in (42) is not straightforward. It is
not simply the case that P in the stative constructions at issue needs verbal
inflection. Rather, we think, that the putative P here has the verb-like property
that it must head a predication to which a -value is assigned. This requires
that this P, like a verb, be situated in a certain structural positionspecifically, it
must head a predicate and it must itself be c-subjacent to a head which sets the
-value of the predicatee.g., T itself, assigning a tense in the traditional sense;
the infinitive to, involved in assigning a dependent or relative tense; or a
causative predicator, like make, which likewise assigns a dependent tense to its
complement (by contrast verbs of the type represented by expect, which assigns
no -value, as is evident from such examples as *we expect John learn Spanish).
This analysis purports to account for the stative readings of certain
experiencer-subject verbs by attributing their stativity to the lexical category of
their heads. By implication it is imagined that the whole business of stativity
might be explained in terms of categoryverbs are active, non-verbs are stative,
to put it simply. Before taking up this issue in more detail, we need to consider
certain problems and consequences related to the basic structural relations
involved in this proposal.
5The
structure depicted in (12) is problematic. Without some special provision, the label
assigned to the upper maximal projection is ambiguousthat is to say, there is no way to
determine which of P and P is the head of the upper projection. We think, however, that the
problem associated with this ambiguity is spurious and that (12) well formed.
6This argument depends, of course, on whether the stative variant of respect the truth can actually
appear in the causative and in the to-infinitive construction of the type shown here. We assume
that the complement in (13a), for example, is stative and that its telic interpretation is due to the
construction; the truly active version, as in respect your parents, means give your parents your
respect, not come to respect your parents. In (13a), the meaning is that an event, or the like, made John
come to respect the truth, not give the truth his respect.
17
(a)
(b)
(a)
(b)
(c)
(d)
7We
have not fully explored the possibility of a Case-theoretic explanation for (14) and the like.
An explanation seeking to limit structural Case to just one internal argument, for example, would
have to explain the range of constructions in which two VP-internal arguments are somehow
licensed without resort to adpositions or other oblique Case morphology (e.g., I envy him his
talent). Such an explanation may well be possible, but we do not pursue it here.
8This is not an autonomous principle, of course, but rather an integral part of the definition of
these two relations, according to which a complement is the unique sister of a head and a
specifier is the unique sister of the first projection (traditionally notated X) of the head. These
notions may ultimately be shown to be wrong, linguistically fictitious, but they are fundamental
to the proposals being entertained here.
18
H1
H1
H2
X H2
H2 Y
(a)
(b)
(c)
(location)
(locatum)
(transitive of inchoative)
If experiencer-subject psych-verbs like fear, respect, envy, etc., also have the
structure in (46), then we must explain why those verbs do not participate in
Middle Formation.
Repeating earlier remarks, the answer to this question, we feel, comes
from the nature of the nominal elements which, by hypothesis, appear in the
lower complement position (Y) in the structures of experiencer-subject verbs.
Consider the expressions cited in (38) and (41), which bear a certain paraphraselike semantic relation to corresponding experiencer-subject verbs. These are
repeated here, in part, as (48) and (49):
(48)
(a)
(b)
(c)
(49)
(a)
(b)
(c)
These all have in common the characteristic that the phrase corresponding to the
emotion, i.e., the psych nominal (my respect, the bosss esteem, their love, etc.),
contains overt material (a genitive nominal or pronominal) representing the
experiencer. Without this (e.g., in Mary has respect, he has love), the character of
these expressions is greatly altered; for all intents and purposes, the experiencer
19
disappears (except to the extent that it can be imagined somehow and variably
attributed).
Importantly, morphology referring to the experiencer in sentences of the
type represented by (48) and (49) is obviative, in the sense that it cannot refer to
the entity corresponding to the closest argument (compare, the similar effect of
the interesting and quite separate semantic principle embodied in the NotionRule of Wechsler, 1995).9 Thus, for example, the genitive pronouns in (50) cannot
be linked to the subject:
(50)
(a)
(b)
And in (51), likewise, the genitive pronouns cannot be linked to the indirect
object, but is linked to the subject (i.e., the more distant argument):
(51)
(a)
(b)
Thus, the psych nominals in such sentences as these contain a genitive which at
once:
(52)
(i)
refers to an experiencer,
(ii)
is obviative, and
(iii) is anaphoric, in the sense that it is necessarily linked to a
c-commanding antecedent if there is one.
(a)
(b)
(c)
(d)
Here, it seems to us, only the general Binding Theory limits the range of
coreference possibilities.
The properties enumerated in (52) essentially boil down to two: the
genitive in psych nominal expressions is obviative and anaphoric. We believe that
this is the key to the problem of the Middle Construction illustrated in (35).
Notice first that in a sentence like (54a), the psych N love, which we assume to
9For
an important recent cross-linguistic analysis of the classical system of obviation, see Aissen,
1997; and for a discussion of an extension of the term to other domains, attributed originally to a
suggestion by Charles Hockett via Joseph Grimes, see Hale, 1992, and references cited there. It is
this extended use of the term which is employed here.
20
give rise to the corresponding verb (through Merge and Conflate), has semantic
properties which are identical to the psych nominal phrase in (54b):
(54)
(a)
(b)
That is to say, the emotion love is attributed to Mary, the experiencer, in both
cases. That emotion is not attributed to the children, whatever the real-world
situation might be. This pattern is true of all experiencer-subject verbs we have
consideredthe conflated noun acts as if it contained a genitive specifier
conforming to the principles of (52). We will assume that something of this
nature is in fact true.
It cannot be literally true that the conflating noun in experiencer-subject
verbs has a genitive specifier, since that would entail that it heads a phrase
(nontrivially) and hence would not conflate with the verb. We will assume
instead that the psych noun (love, respect, envy, etc.) is to be understood as a bare
noun which bears the part relation to some entity (the whole) and, as in
many languages, is related to the latter by means of a relation akin to, perhaps
identical to, secondary predication (as suggested for Part-Whole relations in
Warlpiri, for instance, in Hale, 1981).10 We will employ a bracketed subscript to
represent this informally, and we will speak informally as if the subscript
assigned to the psych noun, in addition to signalling its relation to its antecedent
(bearing the corresponding plain subscript), were an actual item having the
properties set out in (52), specifically the properties of being obviative and
anaphorictechnically, it corresponds to a variable and hence must be bound
(obviatively in these constructions). Accordingly, the abstract structural
configuration given in (46), assuming Y to be the psych noun, would have the
following representation, in which, in accordance with (52), the bracketed
subscript is necessarily disjoint from X, the closest argument, but necessarily
10It
is possible, we suppose, that certain morphological nouns can project a specifier . We have
not considered this to date, but the hypothetical delayed and immediate satisfaction structures (i)
and (ii)corresponding roughly to he died of hepatitus (in an hour), and he was dead of hepatitus (in
an hour) may be relevant to the issue:
(i)
(ii)
D
D
he
he
A
A
of
dead
NP
of
dead
D
t
of hepatitis
NP
hepatitis
It is perhaps more likely, however, that the specifier here (i.e., he) is projected by the preposition
in the standard delayed gratification manner.
21
bound by the next closest argument, the external argument, corresponding to the
experiencer-subject (not shown):
(55)
H1
H1
H2
Xj
H2
H2 Y{i}
Thus, in (54), the subscript is not bound by her children, by virtue of (52ii), but by
the external argument Maryit is Marys emotion, not her childrens:
H
(56)
H
DPj
her children
P
N
P love
{i}
P
DP
the horse
the books
P
N
saddle
shelf
Under ordinary circumstances, a verb with this structure will form a predicate
in sentential syntax and will take an external argument, its subject. The bare
noun will have conflated with the empty P at Merge, and P will have conflated
with V at Merge, and the DP in the internal specifier position will be Caselicensed by the locally c-commanding V.
22
(a)
(a)
(b)
(b)
(c)
(c)
We assume that these verbs have the structure represented in (46) and,
furthermore, that Y is occupied by a noun (the impact noun, e.g., kick, punch,
slap, jab, poke; knee, elbow) which must be linked to its source, the external
argument (i.e., the sentential syntactic subject in sentences like (28a-c), identified
here as the agent role, rather than the experiencer as in the case of the psychverbs). Notationally, the impact noun is supplied with a bracketed subscript, as
11It
should be mentioned, perhaps, that judgments about the Middle are not particularly stable.
With a little thought, most Middles can be made to sound acceptable, or at least imaginable. We
assign stars to middles which require extra thought, recognizing that assessment is relative, in the
sense, for example, that this horse saddles easily is more or less perfect, while this wall kicks easily is
much less than perfect. Interestingly, (b) approaches perfect if the noun punch is taken to refer to
a result or effect, rather than the action attributed to the external argumenti.e., if punch refers to
a dent or depression in the bag, an effect of punching the bag (cf., this bag takes punches
nicely). In this interpretation, punch is more like verbs of the cut-type (see text below).
23
H1
H1
H2
Xi
H2
H2 Y{i}
It follows that these verbs form Middles readily, since the binding requirements
of the result nouns (cut, slice, etc.) and nouns of induced emotion (anger,
fright, etc.) are met internallythus, in (60), for example, the separation in
material integrity entailed by a successful instance of cutting, slicing, and the
like, is an acquired property of the internal argument (Xi), not of the external
argument, i.e., of the sentential syntactic subject in the transitive; similarly for
experiencer object verbs, the induced emotion is linked to the internal
argument:
(60)
(a)
(a)
(b)
(b)
(c)
(c)
(d)
(d)
(e)
(e)
Assuming that these remarks about structure are correct and, specifically,
that it is correct to assume that the experiencer-subject psych-verbs share the
same structural configuration with location and locatum verbs, we can assume
further that the stative and nonstative uses of verbs like respect, love, and so on,
are identical structurally. In this instance, at least, stativity does not correlate
with structure, in the sense of syntactic configuration, but with something else.
By assigning to stative experiencer-subject psych-verbs the representation
in (42), we have forced the issue, claiming that their statitivity is a matter of
category, with V nonstative and P stative. But this is an artifice, a trick designed
24
to make category and stativity coincide. Moreover, we have not investigated the
consequences of the kind of head movement invented here to derive the
structural configuration in (42), and we have probably violated the principles
underlying the relations involved in the extended projections which define
sentential syntactic constructions (Grimshaw, 1991), principles strongly
suggested by our intuition that the very definition of the category V is the
morphological one according to which a verb takes tense and aspectual
morphology.12
3.2. True stative verbs.
To say that experiencer-subject verbs of the kind exemplified in (36) [fear,
knew, admired, liked, respected] are stative is probably inaccurate. This is
suggested both by the fact that they are open to non-stative interpretations in
appropriate contexts and by the findings documented in a rich body of literature
on aspect which provides copious demonstration of the fact that stativity, telicity,
and the aspectual classes (activities, accomplishments, achievements), pertain not
to verbs but to the predicates they head (cf., Dowty, 1979, 1991; Tenny, 1987,
1992). It would be reasonable to entertain the possibility that these notions, and
stativity in particular, are never features of individual lexical itemse.g., of
verbs, nouns, adjectives, adpositions, or what have youbut rather of whole
predicates.
But this does not seem altogether satisfactory either, for some heads are
entirely consistent in their behavior in relation to so-called stativity. For example,
the functional head (covert or overt) defining the extended projection of the
category adjective is consistently stative. Thus, while the verb phrase turn greener
is nonstative, this is a property of the verb phrase headed by turn; the adjectival
extended projection headed by -er (putative category ) is itself stative (as it is
in (33b) and (35c) above), a property evidently attributable to the functional
head.
The category V is not entirely left out here, since some verbs head
predicates which are classically stative:
12Systems
of the type represented by Hopi (Jeanne, 1978), in which tense and aspect morphology
selects the category P, as well as V, may or may not counterexemplify the principles of Extended
Projection. This will depend on a variety of factors. In the related Oodham, for example, the categories N and A take tense and aspect morphology, superficially, but it can be argued that these
cases involve incorporation of bare nominal and adjectival stems into a morphophonologically
suffixal copula -k(a) derived from the Uto-Aztecan verb *kat sit, be. It is this copula which takes
tense and aspect morphology, not N and A directly. The case is not as simple as this for Hopi,
inasmuch as, if there is a copula there, it is not overt and its detections will require more work. In
general, however, the principles of Extended Projection are supported empirically to an extent
which encourages us to assume that the Hopi system will eventually be shown to fall in with the
general case.
25
(61)
(a)
(b)
(c)
(d)
These are stative in much the same way copular sentences with be are stative:
(62)
(a)
(b)
(c)
(d)
Furthermore, if we take the position that the verbs of (61) are in reality copulas,
sharing certain essential properties with the copula be, then their most renown
property can be explainednamely, their failure to participate in the passive
construction:13
(63)
(a)
(b)
(c)
(d)
Suppose that the verbs of (61) are copulas, in fact, differing from be by
virtue of their lexical (as opposed to functional) status and correspondingly
richer semantic content, sometimes paraphrasable by means of a prepositional
modifier, as in (64a, b):
(64)
(a)
(b)
Under this interpretation, the verbs of (61) do not select an object complement,
but rather a predicate, as pointed out often in the literature on these topics. Thus,
while the expression fifty thousand dollars is a standard (plural) object DP in the
passivizable (65a) below, it is a predicate in the unpassivizable (65c) (cf. (61a) and
(63a) above):
13There
is an important property of the copula be which is not shared by the semantically more
contentful verbs of (31). Even in its copular function, be behaves like an auxiliary in relarion to
inversion (I-to-C raising)e.g., Is two and two four?
Some of the verbs of (33) can passivize, of course, in a different use. And, (33c-d)
themselves are weakly possible, using equal and comprise in senses somewhat different from those
attributed to them inthe suggested copular use. The well-formedness of the passive verb form in
the collection is comprised of three books is a different issue. In general, measure phrases of the type
found in (33) sound rather bad as subjects of passivese.g., ??$5 was earned by John. This cannot
account for (33), however, since in the corresponding Wh-questions, the passive is possible with
earn, as in how much is earned by each worker, while with cost, for example, it remains ill-formed, as
in *how musch is cost by that house?
26
(65)
(a)
(b)
(c)
The lexical head which projects the clause in this casei.e., worthis nominal in
category, requiring support by the auxiliary be, as expected. But it is
syntactically a copula, and its structural complement, the measure phrase, is a
predicate and not the sort of complement which is expected to be case marked by
the head that selects it. Hence, of-insertion (which is otherwise required, as in the
worth of her suggestion) is not applicable.
Although the details are far from clear, it is possible that a similar analysis
is appropriate to another class of verbs which fail to passivize (cf., Perlmutter
and Postal, 1984:92):
(67)
(a)
(b)
27
(if at all), given that the two evidently select identical complements (here,
yellow)?
(68)
(a)
(b)
(a)
(b)
(70)
(a)
*With Annan to Bagdad, we can relax.
(b)
*With Kirsten from Lincoln Center, New York will boycott
the ballet.
(a)
(b)
The plain prepositions in and on can express a relation in which the argument in
Specifier position (i.e., derived s-structure subject) corresponds to an entity
which moves or is arrayed along a path ending at the place denoted by the
complement, like the related prepositions into and onto:
(72)
(a)
(b)
But this is not the reading which comes through in the with-construction
exemplified in (71). Instead, in those examples, the understanding is that the
location of the entity denoted by the Specifier in the P-projection coincides in a
certain sense with the place denoted by the complement.
The opposition which emerges in (69) and (70) is one which appears to be
rather pervasive in the lexical and functional systems of the grammars of natural
languages. It is probably to be identified with the well known telicity opposition,
and with the central and terminal coincidence opposition to which we have
referred on occasion (cf. Hale, 1986). The prepositions of (69) project the dyadic
structure characteristic of the lexical category P:
(73)
P
XP
P
P YP
We have suggested three nuclear types which are inherently stative in this
sense: (i) the head which defines the extended projection of A, i.e., the category
(as in (32, 33) above); (ii) a subclass of the category P, e.g., in, at, as in (69a, b); (iii)
the copula, morphologically a subclass of V, e.g., cost, weigh, as in (61).
The first of these can be illustrated by means of the small clause in (32a),
repeated here as (74), with structural representation in (75):
(44)
(45)
DP
the sky
clear
This is claimed to involve central coincidence because its specifier, the sky,
corresponds to an entity which possesses the attribute denoted by the
complement, i.e., the adjective phrase clear. That is to say, the relation between
the specifier and the complement is not one of change. The entity denoted by the
specifier possesses the attribute. It does not come to have the attribute, or come
to lack the attribute, but rather, the entity and the attribute coincide to define a
set whose members are at once the sky and clear. Contrast (75) with (77) below,
corresponding to the inchoative, i.e., terminal coincidence, hence nonstative, (76):
(76)
(77)
V
DP
the sky
V
A
V clear
The dyadic head V, like the majority of verbs, has the property of projecting a
structure expressing the terminal coincidence relation. The entity denoted by the
specifier undergoes a change whose end point is possession of the attribute
denoted by the complement.
Central coincidence prepositions, like in in (69a), repeated here as (78),
project a wide variety of structures showing a correspondingly wide range of
interpretations. In this case, the preposition is used to express its customary
locational sense and function:
30
(78)
(79)
P
DP
Annan
P NP
Baghdad
in
The entity denoted by the specifier, Annan, coincides with the location denoted
by the complement, Baghdad. Here again, no change is expressed in the small
clause. Rather, the preposition identifies the location of the entity denoted by the
specifier with the place denoted by the complementthe two locations coincide
centrally, not terminally, in so far as that is physically possible. By contrast, in
they led Annan into Baghdad, the preposition expresses terminal coincidence (the
place, Baghdad, being the terminus ad quem).
Turning now to the stative copula, we believe that central coincidence is
what defines that category of verbs. In a predication of the type represented by
(80), employing the prototypical copula be, the property denoted by the syntactic
complement, i.e., the predicate nominal a calf roper, is attributed to the entity
denoted by the subject:
(80)
In this case, the predicate nominal denotes a property which corresponds to the
end point of a change undergone by the entity denoted by the subjecta relation
comparable to that in (76) above, and unlike that in (78), which is to be compared
rather with (80). The verbs be and become, in (80) and (81), constitute a minimal
pair, so to speak, for the central versus terminal coincidence opposition.
Our conclusion about stativity is that it is not itself a feature of heads.
Rather, it is a property of constructions and arises in the semantic composition of
meaningful elements. However, among the elements which contribute to a
stative semantics is an element which is attributable to syntactic heads. This is
the semantic opposition just discussed, i.e., coincidence. Some heads must be
identified with central coincidence. Among these are some verbs. The stative
copulas (e.g., be, cost, weigh, equal) are clearly members of this class. We leave
open the question of how widely central coincidence is distributed among the
rest of the verbal lexicon.
31
(83)
P
DP
the baby
P
P
in NP
bed
(85)
P
DP
the baby
P
to
P
P
in NP
bed
(a)
(b)
(a)
(b)
V
DP
DP
Leecil
V
stay
Leecil
P
P N
go
in Tucson
to
(at) Tucson
The verbs stay and go select the P-projections indicated. They are not themselves
central or terminal coincidence. That property derives from the P-projections.
Accordingly, the verbs are not necessarily stative. Any stativity which might
adhere to these sentences is due to the P-projections, and it correlates with the
central and terminal coincidence distinction inherent in the configurations. The
simple P-projection (as in (87a) corresponds to central coincidence, and the more
complex structure (P within P, as in (87b)) corresponds to terminal coincidence.
Transitive counterparts to (86) are illustrated in (88) and (89):
(88)
(a)
(b)
(89)
(a)
V
V
keep
DN
the calves
put
P
DP
DP
the calves
in the corral
to
P
DP
With these examples we claim that a genuine opposition exists between two
kinds of verbs, depending on the type of P-projection which appears in their
lexical structures. Central coincidence verbs are those built upon a simple
P-projection. By contrast, terminal coincidence verbs are built upon complex Pprojections; they contain a P-projection consisting of a P which takes a second P
as its complement.
33
4. Semeliterative re-.
Finally, we turn to a prominent feature of English morphology which
belongs clearly to the realm of aspect, in the sense of this conference. Our interest
in this element, i.e., semeliterative re-, is more syntactic than semantic, but like
other examples discussed here, it represents the semantics-syntax interface. We
are concerned here with the phenomenon illustrated by the following paradigm,
based on Keyser and Roeper (1992):
(90)
(a)
(b)
(c)
(d)
(e)
(f)
position. It is this trace, then, that fills the clitic position, to the exclusion of any
other like element. Taking the prefixes re- and over- to be like elements, they
are excluded from co-occurrence with the particle up, because the trace of the
latter occupies the unique clitic position. Similarly, the occurrence of two
particles construed with a single verb, as in (92c,d), is likewise impossible, since
there is room for only one trace.14
The proposal which we will make here is spiritually akin to the analysis
developed by Keyser and Roeper . In particular, it exploits the fact, amply
documented in that study, to the effect that there is somehow room for just one
element in the outer morphology of the verb. In our framework, however, we
are constrained in the matter of a clitic position. The idea that there is a unique
position for clitics, assuming that the class of clitics is itself definable, is not a
notion which has a place in the theory of argument structure which we are
considering (cf., Hale and Keyser, 1993a,b).
Let us take a clitic to be an adjoined head. Since there is, in principle, no
limit to the number of possible head-adjunctions in the formation of a complex
verb word, there can, in principle, be no single clitic position. Some other factor
must be at work in determining the mutually exclusive distribution of the
elements under study here.
4.1. Non-overt adjuncts, the witchs heart
One observation which casts doubt on the single-clitic hypothesis is exemplified
by the verb forms in (93) below:
(93)
These are far from elegant, to be sure, but they seem perfectly well-formed to us.
They are not at all in the category of (90e,f) or (92c,d), for example. If re- and overare clitics, in the sense of Keyser and Roeper, then their co-occurrence is in
defiance of the single clitic hypothesis, clearly.
While the prefixal clitics are not in complementary distribution with each
other, they are in complementary distribution with the particles. This rings a bell,
in fact. If, as we have suggested, a particle is part of a chain, with a trace located
14While
the exclusion of two or more particles might be a semantic matter., as suggested in the
text, it is also a syntactic fact, one which the Keyser and Roeper analysis accounts for
straightforwardly. A semantic analysis might appeal to the idea that multiple clitics would be
expected to correspond to multiple contrastive scope relations; the fact particles do not co-occur,
might be put down to some scope competition, or some such thing. But this would simply be
the result of the syntactic mutual exclusion, so far as we can see. It should be mentioned further
that particles do co-occur, as in take the pot back off (the stove). From our point of view, however,
sequences like back off X form single constituents and do not counterexemplify the claim that
particles are mutually exclusive; in the example just given, the verb take selects a single particle
(complex). We also exclude from consideration here coordinated particles, as in he bounced up and
down.
35
in the verb word, then the problem may well have to do with that very fact. That
is to say, perhaps it is not the particle itself but rather the associated non-overt
element that is the cause of the ill-formedness. This brings to mind the
impressive body of evidence marshaled by Pesetsky (1993) showing that
limitations on non-overt morphology are fundamental in characterizing and
explaining a wide range of ill-formed structures. These limitations are succinctly
expressed in Myers (1984), in the form of the generalization now widely known
by its authors name:
(94)
Myers Generalization:
Zero-derived words do not permit the affixation of further
derivational morphemes.
The effect of this is to rule out any derived word having the following abstract
form, where X, Y, and are heads, and the manner of composition is by
conventional adjunction (order of head and adjuncts irrelevant):
(95)
[X Y [X [ X ] ] ]15
Let us assume that (96a,b) involve the attempt to nominalize the verb give in the
so-called double object construction, to which we have attributed a structure
essentially like that associated with locatum verbs, of the type represented by
saddle, clothe, etc. (cf. Hale and Keyser, 1993a). In our structure, the verb give has
the function of lexical constant, as in the following:
(97)
V'
V
VP
V'
NP
V
give
PP
P NP
36
[V [V]]
This raises and substitutes for the upper V, giving the final form of the transitive
verb. The derived upper V, a product of substitution, not adjunction, presents the
same configuration as (98)it is not, therefore, in violation of Myers
Generalization. Further derivational morphology, however, would give rise to a
violation, as in the nominalization:
(99)
N'
N
V'
-t
VP
V'
NP
V
give
PP
P NP
Here, the derivation of the verbal subpart proceeds as for (97) above; but the
final adjunction (to N) gives rise to a configuration which is in clear violation of
Myers Generalization:
(100) [N [V [V]] -t]
Hence the ill-formedness of (96a,b). By contrast, the nominalization formed from
the to-Dative construction, since it does not involve a phonologically null
adjunct, does not obtrude Myers Generalization:
(101) Johns gift (of a book) to Bill
It is tempting to attribute the behavior of the prefixes re- and over- to the
principle underlying Myers Generalization.16 In this connection, consider the
following set, modeled on examples in Keyser and Roeper (1992:90; compare
especially their (91h)):
(102) (a) I sent the letter to them.
(b) I resent the letter to them.
(c) I sent them the letter.
16We
have not attempted to determine what this principle might be, but we suspect that it is the
requirement that all terminal nodes be fully interpreted. The witchs heart may simply be
invisible for interpretation, presumably at LF.
37
38
We are now left with the residue of the uniqueness, or single position
hypothesis.
4.2. Uniqueness
We think that the proper understanding of examples like (90e,f), (91b), and
(92c,d) does in fact lie in uniqueness. There is something right about the idea that
there is no room for both re- and a particle, or for more than one particle. The
problem is to determine the precise nature and location of the restriction. We
have rejected the single clitic hypothesis on the grounds that there is no nonstipulative way to limit adjunction in word formation, assuming that clitics are
simply adjoined morphemes (heads). Hence, if the apparent uniqueness
restriction is to be explained, it must fall out from general properties of
grammar.
Let us approach the problem by first considering how the verbs of (105a-c)
are derived:
(105) (a) We heated the soup.
(b) We heated the soup up.
(c) The soup heated (slowly).
(d) The soup heated up (finally).
Our assumption to date has been that verbs of the type represented by (105a) and
(105c) are the result of head-movement from the adjectival complement of an
abstract verb V, giving the intransitive (105c). Subsequent movement to a higher
V yields the transitive (105a). This has been our basic position in relation the
derivation of de-adjectival verbs. An essential feature of this simple view is that
the adjective raises directly out of the AP, and the latter is the immediate sister of
V, as shown in the structural diagram (106):
(106)
V'
V AP
A
In this respect, de-adjectival verbs differ from denominal verbs of the location
and locatum types, like shelve, corral, saddle, bridle the latter involve raising of N
first to P and then to V (cf. Hale and Keyser, 1993a). While the adjective reaches
the verb in a single step, the location or locatum noun reaches the verb in two
steps.17
17This
represents the account of Hale and Keyser (1993); our current view, of course, is that the
relations holding in the location and locatum verbs is the relation of "selection". The analogue of a
two-step composition still holds, however. This being the case, we will speak in terms of Hale
and Keyser (1993), bearing in mind that the derivation of denominal verbs does not involve
incorporation. The derivation of de-adjectival verbs evidently proceeds via incorporation (as in
[Vredd[Ven]]).
39
AP
A
An empty will, by general assumption, require interpretation (at PF). This is
accomplished, let us suppose, by head-movement. The adjective raises first to ,
thereby satisfying the Full Interpretation requirement for ; the latter then raises
to V, satisfying the Full Interpretation requirement for that empty category as
well.18
We are proposing, in effect, that an adjective phrase in these constructions
is always dominated by a lexical category . This has a semantic correlate, we
assumeit is the component of an adjectival expression which refers to the
degree or intensity at which the quality denoted by the adjective is realized.
We maintain that this is basically correct. If it is indeed correct, then we
have a source for the particles up, off, down, and the like, as they occur in
association with de-adjectival verbs. These particles represent the category , and
they arise, we contend, in the same way that cognate and hyponymous objects
arise (cf. Hale and Keyser, 1993b).
In an earlier analys, a hyponymous object construction, like dance a new
dance or dance a jig, involved insertion of a DP in the position corresponding to
the complement of the verb dance. However, this position was in fact the point of
origin of a chain headed by the noun dance, adjoined to the abstract V heading
the verbal projection as a whole. Thus, head-movement was assumed to derive
the structure depicted in (108), where the point of origin, or foot, of the chain
corresponds to Ni1 , the head of the chain being the adjoined Ni2 :
(108)
V'
V
NP
N i2 V N i1
18The
40
19Jon
Bobaljik, in his Notes From Kamchatka (e-mail, 1994), considers a theory of cognate objects
according to which they are inserted in substitution of traces at the foots of chains defined by
incorporation. He develops this idea in the context of a typological account of the variability
observed within the class of so-called unergative verbs. Our own view of this matter is not
sufficiently well articulated to determine whether we agree or disagree with this, but both
Bobaljiks proposal and ours share the conviction that the classificatory nature of the
putatively incorporated material is properly expressed by positing a chain whose foot is
associated with certain classificatory featuresfor us these are associated with the trace and its
projection, for Bobaljik they are associated with the index and the category originally dominating
the trace.
41
(110)
V'
V
P
AP
A
Thus, when the adjective raises, first to and then to V, we are left with traces in
A and . The head of the complement is , of course, and the category of the
complement is as well. We might suggest that when reinsertion takes place,
giving the equivalent of a cognate argument construction, it is the category , not
A, that is inserted. This is the source of de-adjectival verb-particle constructions
like those of (111):
(111) (a) heat up
(b) cool down
(c) cool off
(d) widen out
The particles here are (intransitive) members of the category . By hypothesis,
then, the expressions in (111) are adjectival correlates of the more familiar
cognate argument constructions of English and other languages.
Under this assumption, the ill-formedness of (92c,d)i.e., *cool off down,
*cool down offis explained as a function of the principle according to which the
head-complement relation is biunique. The category V has the essential property
that it takes a (single) complement and forms an expression denoting an event
type. There is, therefore, room for only one complement, and since the
particles off and down represent the complements of the verb cool, there is room
for only one particle, a condition which is met in the well-formed (92a,b), cool off,
cool down.
This analysis extends as well to (90e,f) and (91b)*reheat up, *overheat
upunder the assumption that the prefixes re- and over- are members of the
category , originating in the complement of the verb to which they are attached
at s-structure. The prefix and the particle are mutually exclusive for the same
reason that particles are themselves mutually exclusive, namely, the principle of
the uniqueness of complements.
To the extent that this suggestion is based on a true linguistic
principlethe uniqueness of the complement relation it has the quality of a
real explanation for the mutually exclusive distribution of prefixes and particles.
But there are some questions remaining. We have not really explained the illformedness of (109a,b), assuming that they are in fact ill-formed. Our explanation
will have to be tied to the category . We must assume that this category is
always present in the structure which gives rise to de-adjectival verbs of the type
being considered here. That is, the basic structure of the verb heat has the form
depicted in (107), necessarily. A simple adjectival (AP) complement is not a
possibility there; if it were, then heat too hot should be perfect as a cognate
42
assume, of course, that heat too hot is ill-formed as a cognate complement construction.
However, judgments vary considerably in this regard. The expression redden vermilion for
example, is judged worse than heat too hot. The latter might involve adjunction in sentential
syntax, rather than complementationcompare heat up too hot, in which too hot must be an
adjunct, by hypothesis. In this connection, see also the text in relation to (23) below.
43
In both uses, the particle forms a constituent with the PP which follows it,
assuming that its pied-piping potential can be construed as indicative of that.
And the two uses also share the property that they can be modified by right:
(116) (a) cool right down (to 0, to the lowest temperature possible)
(b) heat right up (to 1000, the the highest temperature)
(c) drop right down (to the bottom)
(d) rise right up (to the top)
Thus, the particles up and down give every indication of being the same thing in
the two observationally different uses seen here. They share the property that
they may take an overt complement, if that is a PP.21 An overt adjective cannot
appear as the complement of up or down, as shown by (112) and (113) above. We
have claimed, however, that whileup and down cannot take an adjectival
complement, there is an empty member of the same category which can, namely
the element which we have symbolized . Presumably, this null variant lacks the
morphological properties which compel its overt brethren to select a particular
category. We will assume that this is true and proceed to identify this property,
and thereby the category of itself.
We are assuming, of course, that the null element and the overt particles
up, down, etc., belong to the same category. This being the case, let us temporarily
refer to the entire class as . And let us assume that overt is parasitic on P, in
the sense that it inherits a certain property from Pit can take an overt
complement only if P supports it. The property in question might well be one
which is relevant in sentential syntax. In fact, it is quite likely that Case is at issue
here. This is consistent with the fact that can take a nominal complement, but
not an adjectival one. Of course, can assign case only if P, a true case
assigner, intervenes. We suggest, therefore, that is a kind of Pa defective P,
let us say. When it is overt, and when it takes an overt nominal complement, it
must combine with a full preposition in order to transmit Case to the nominal,
thereby satisfying a sentential syntactic requirement (the Case Filter).
This proposal assumes that there exist Lexical Relational Structures of the
following abstract form:
(117)
V*
V
PP
P 1 PP
P 2 NP
is possible, of course, that up and down also take overt DP complements, as in up the tree, down
the road. We are not decided on the question of whether these usages are the same as that
represented by (25) and (26).
44
is not surprising, therefore, to learn that adjectival predicates in some languages are regularly
marked for some a category relating to the presence, or absence, of the condition denotedstage
level adjectives are so marked in Oodham (Pima-Papago) for example.
45
and the same predicate. And this, like any adjectival predicate, finds its subject
external to its own projection. Accordingly, its subject (the soup, in the example
at hand) appears in the specifier of the immediately superordinate verb, V2 . The
substructure headed by this verb could itself have constituted the matrix VP,
giving the intransitive (the soup heated). But (119a,b) represent the transitive
variant corresponding to the structure the depicted in (120):
(120)
V*
V1
VP
NP
N
soup
V'
V2
PP
P AP
A
The derivation of (119a) proceeds as follows. The P here is empty and therefore
must be given phonological constituency in order to satisfy the Full
Interpretation requirement at PF. As usual, this requirement is met by
incorporating the head of its complement, here the adjective, A. This process
effects a merger of the phonological matrix of the overt A with that of the nonovert P. The governing verb, V2, being phonologically null, must also incorporate
the head of its complement (the newly derived [ P A-P]), and the uppermost verb,
V1 , must likewise incorporate the head of its complement (the derived complex
head [V [PA-P]-V]). Given the English pattern of incorporation exemplified here,
the phonological representation of the derived matrix V1 is traceable ultimately
to the overt source, ordinarily the most deeply embedded lexical head involved
in the sequence of incorporationsin the present example, it is the adjective hot
in its derived verbal form heat.23
Now consider (119b), with the particle up. The derivation of the verb and
VP proceeds by incorporation, in the same manner as for (119a). By hypothesis,
after incorporation, there is a trace corresponding to the PPmore specifically,
there are traces corresponding to A and P. These are not referential traces, but
rather traces of lexical heads. They represent positions in which lexical heads, of
appropriate categories, can appear. We may suppose that these traces are
nothing more than empty terminals in the syntactic structure. If this is correct,
then it is possible that the positions which they occupy are positions into which
lexical material can be reinserted. This is what we have proposed in order to
account for the phenomenon of cognate complements (see above, and see also
Hale and Keyser 1993b). And this is what we propose for (119b). The overt
particle up is inserted in place of the empty P, i.e., in the position corresponding
23To
be sure, it is unlikely that there is a synchronically real phonological derivation here; the
relationship is essentially suppletive, but in many othercases, e.g., warm, cool, the adjective and
the derived verb are homophonous, and in many, the relation involves affixal morphology, as in
redden, widen, darken.
46
24It
is important to determine that this is a true principle and not a stipulation. We are not
prepared to argue this point here, but we imagine that the principle is inherent in the projection of
syntax from the lexicon and the general requirement that projected structural relations are
asymmetrical in the sense of Kayne (1993).
47
25We
should point out, perhaps, that uniqueness itself is not strictly speaking an autonomous
principle of grammar. Rather it is a consequence of a principle which might be termed the
asymmetry of projections (cf. the antisymmetry of Kayne, 1993). Suppose this principle requires
that: if A c-commands B, where A and B are at the same level of projection (X, X, XP), then B
does not c-command A. From this it will follow that all branchings are binary. This principle
might itself be the consequence of a deeper linguistic principle, to wit Full Interpretation (cf.,
Chomsky 1986).
26Tom Roeper has opointed out to us that the verb reenter is ambiguous in the manner being
suggested here. The sentence he reentered the atmosphere. can mean (a) he entered twice or (b) he
came back into the atmosphere (only one entry). This seems to be general for initial/new states
(enter, turn out, insert, etc.) but not for gradations (warm, heat, cool); with re-, and without
further specification these latter verbs, it seems to us, have only the back to some level, or
narrow scope re-, reading.
48
49
V'
V AP
A
This is, at the very least, a recognizable intuition about this type of illformedness. The structure shared by (132a,b) is presumably that shown in (133),
being the position assumed for the prefic re- and the particle up:
(133)
V
V
DP
soup
warm
A
This position can be "filled," of course, by just one element, in accordance with
(130). Hence (132a) or (132b) is possible, but (131) is not.
5. Concluding remarks and observations.
What is the relationsip between aspect and argument structure? The
question makes sense, of course, if the terms are defined. We define argument
structure as the system of structural relations holding between lexical heads
27The
biuniqueness of grammatical relations may follow ultimately from full interpretation, but
the two notions are conceptually distinct, and it would be premature to conclude, for example,
that the principle of the asymmetry of projections is itself redundant.
51
(nuclei) and their arguments within the syntactic structures projected by nuclear
items.
Our conclusion, in general, is that aspect is orthogonal to argument
structure. Whenever we deal with questions of interface and interaction in this
domain, we observe that argument structure is for the most part autonomous. Its
properties and characteristics are strictly local, being defined in terms of the
structural relations of complement and specifier. To be sure, any argument
structure configuration associated with an actual predicate in sentential syntax
will be interpreted in terms of one or another aspectual type (achievement,
accomplishment, etc.) and its arguments will be associated with one or another
aspectual role (measure, path, terminus, etc. (Tenny, 1994)). But argument
structure is a distinct and separate component of grammar.
52
References
Ackema, Peter, and Maaike Schoorlemmer. 1995. Middles and Nonmovement,
Linguistic Inquiry 26:173-197.
Basilico, David. 1998. Object position and predication forms. NLLT 16:541-595.
53Bittner, Maria, and Ken Hale. 1996. The structural determination of Case and
Agreement. Linguistic Inquiry. 27.1-68.
Chomsky, N. 1986. Knowledge of Language. Praeger.
Condoravdi, C. 1989. The Middle Voice: Where Semantics and Morphology
Meet, Student Conference in Linguistics 1989. MIT Working Papers in
Linguistics 11. Cambridge, MA: MIT.
Fagan, Sarah. (1988) The English Middle, Linguistic Inquiry 19:181-203.
Fagan, Sarah. 1992. The Syntax and Semantics of Middle Constructions. Cambridge:
Cambridge University Press.
Hale, K., and J. Keyser. 1993a. On Argument Structure and the Lexical
Expression of Syntactic Relations. In The View from Building 20: Essays in
Linguistics In Honor of Sylvain Bromberger. MIT Press. Pp.53-109.
----------, and ----------. 1993b. The Complex Nature of Simple Predicators. To
appear in Alex Alsina et al. (eds,) Proceedings of the Stanford Conference in
Complex Predicates. CSLI.
Hale, Kenneth and Jay Keyser. 1998. "The basic elements of argument structure."
Pp. 73-118 in Papers from the UPenn/MIT Roundtable on Argument
Structure and Aspect. MIT Working Papers in Linguistics, Vol. 32.
Cambridge, MA: MITWPL.
Hale, Kenneth and Jay Keyser. 2000a. Conflation. In Ana Bravo Martn, Carlos
Lujn Berenguel, Isabel Prez Jimnez (eds.) Cuadernos de Lingstica VII
2000, Documentos de Trabajo. Lingstica Terica. Madrid: Instituto
Universitario Ortega y Gasset. Pp. 39-76.
Hale, Kenneth and Jay Keyser. 2000b. "On the time of Merge." MIT ms.
Jackendoff, Ray, 1983. Semantics and Cognition. Cambridge, MA:MIT Press.
Jackendoff, R. 1990. Semantic Structures. MIT Press.
Kayne, R. 1993. Antisymmetry. CUNY Manuscript.
Keyser, Jay, and Tom Roeper. 1984. On the Middle and Ergative Constructions
in English, Linguistic Inquiry 15:381-416.
53
54