Adger and Tsoulas- Circumstantial Adverbs and Aspect

download Adger and Tsoulas- Circumstantial Adverbs and Aspect

of 25

Transcript of Adger and Tsoulas- Circumstantial Adverbs and Aspect

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    1/25

    Circumstantial Adverbs and Aspect1

    David Adger George TsoulasQueen Mary University of York

    University of London

    1 The syntax and semantics of manners and locatives

    In this paper we focus on the syntactic licensing and structural position

    of manner adverbials (specifically those in lyin English) and locatives. We

    show that there is a correlation between the licensing of arguments and the

    licensing of these elements. We argue that the observed correlations can becaptured by developing a theory of the licensing abilities of a series of

    functional heads in the verb phrase.

    It has been known since Harris (1968) that there is a correlation between

    the ability of certain verbs to take a manner adverbial and their ability to

    subcategorise for an affected object. Thus, verbs like resemble, have, cost

    etc. are ill formed with manner adverbs:

    (1) *John resembled Sue slowly.

    (2) *John had flu worriedly.2

    (3) *The slave cost 600 denarii wholeheartedly

    These verbs, although they take an object which is apparently marked

    with accusative case, do not allow of-insertion in their nominalisations3:

    (4) a. *John's resembling of Sue

    b. *John's having of flue

    c. *The book's costing of 30

    We can take this as showing that these verbs do not mark their objects

    with true structural accusative, but rather with inherent accusative case (see

    Torrego 1998 for evidence that accusative can be inherent).A fact which is linked to this is Chomsky's (1965) observation that

    certain verbs which do not passivise, also cannot occur with manner

    adverbials, a fact which he explained by allowing a passive-triggering

    morpheme to be generated in the subcategorised manner position:

    (5) a. *Sue was resembled by John

    b. *Flu was had by Maria

    c. *50 were cost byRhyme and Reason.

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    2/25

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    3/25

    3

    A striking correlation between locatives and telicity is noted by Borer

    (1998b). She shows that, in Hebrew, certain inversion processes are

    sensitive to the telicity of the predicate, so that inversion is impossible withatelic verbs, giving the contrast between the (a) and (b) examples.

    (18) a. Parcu mehuma (telic)

    Erupted riot

    "A riot erupted"

    b. *avad ganan (atelic)

    Worked gardener

    "A gardener worked"

    Note that with the atelic predicate in (18b) inversion is impossible.

    Interestingly, when a locative element is inserted, inversion becomes

    possible:

    (18) c. avad sam/kan/ecli ganan

    Worked here/there/chez-me gardener

    "A gardener worked here/there/at my house"

    Again, what we seem to have here is a close link between the expression

    of a locative and the aktionsart of the predicate, which may then have further

    syntactic effects. Of course, all of these constructions involve other

    interfering factors (sentence aspect, focus/presupposition structure) and it is

    not our intention to provide an analysis of all correlations between locatives

    and telicity. What we will try to do is outline a framework which is capable

    of making the necessary links, and show in a couple of core cases how the

    analytic potential of the system plays out.

    In this paper, then, we seek to give a unified account of the way that

    manner and locative adverbials interact with case and aspect specifications.

    In doing so, we are led to a particular view of these low-VP adverbials which

    essentially places them within the low functional structure of VP, and where

    their licensing takes place through featural mechanisms similar to themechanisms which are thought to be involved in the licensing of true DP

    arguments.

    2. Some further data

    Many accounts of manners and locatives place them low within the VP,

    with locative adverbials hierarchically superior (Andrews 1982, Bowers

    1993, Ernst 2001). This approach can make sense of the fact that manner

    adverbials appear closer to the verb than locatives:

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    4/25

    4

    (19) We tortured the general slowly in the garden

    (20) ?We tortured the general in the garden slowly

    Example (20) is only well formed with a clear prosodic break between

    the two adverbial phrases, suggesting that the manner adverbial has been

    'moved' rightwards (possibly in the prosodic component of the grammar -

    Zubizarretta 1998), or that, at least, it is not in its canonical position.

    One traditional approach would be to assume that the manner adverbial

    is generated lower down than the locative, and both are right adjoined to

    some projection of the verb (21). This data then follows naturally.

    (21) [ [ [ V ] Manner] Locative]

    This type of approach also captures the fact that manner adverbials are

    far more restricted in their syntactic distribution than, say, temporals:

    (22) This bridge may (*badly) have (*badly) been (badly) designed

    (badly) by Brunel (badly).

    (23) *We slowly often tortured the general

    (24) We often slowly tortured the general

    Note that, in (22), the manner adverbial is restricted to positions very

    close to the verbal predicate itself, or to rightward positions which again

    may be assumed to arise because of prosodic factors. Similarly, (23) and (24)

    show that, while a manner adverb may appear to the left of the verb, it may

    only do so if it is adjacent to it.

    It is also well known that preverbal manner adverbials have a different

    interpretation in English from postverbal ones (Stalnaker & Thomason

    1973):

    (25) a. He has been slowly testing some bulbs

    b. He has been testing some bulbs slowly

    In (25b.), each particular test must be slow, whereas this is not the casein (25a.).

    One approach to this might be to assume that the preverbal adverb is

    actually adjoined to the lexical verb itself, rather than to some higher

    projection, giving the following (rough) structures:

    (26) been [slowly testing] some bulbs

    (27) been [testing some bulbs] slowly

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    5/25

    5

    In (26), the adverb modifies the event denoted by the verb directly, with

    the semantic consequence that the testing event is slow. In (27), the adverb

    modifies the VP, including the object so that each event of bulb-testing isslow. We are not wedded to this analysis, but we do assume that there is

    something syntactically special about pre-verbal manner adverbs, and we

    concentrate in our discussion on post-verbal ones.

    One further argument for the position of post-verbal manner adverbs can

    be constructed on the basis of Costas 1996 discussion of adverb placement.

    Costa argues that PP arguments which occur after manner adverbs are in

    their base position, since they do not show any of the freezing effects one

    would expect of PPs in extraposed positions:

    (28) Which woman did he glance quickly at a picture of t?

    (29) *Which woman did he glance yesterday at a picture of t?

    compare the sentence with a locative:

    (30) *Which woman did he glance in the Louvre at a picture of t?

    This paradigm suggests that locatives are in a zone of the sentence after

    which extraposed elements appear, whereas manners are not (or, at least, do

    not have to be). Note, again, however, that this does not push us into an

    analysis where locatives are generated higher than manners, since the

    extraposition zone may actually be rather low down (see, for example,

    Haider 1998, or Kayne 1994).

    Two final points to note in this discussion of the relationship between

    manners and locatives is, first, that that PP manner adverbials induce the

    same effect as locatives and temporals (31), and, second, that these manner

    phrases may not occur preverbally without comma intonation (32):

    (31) *Which woman did he glance in a sultry way at a picture of t?

    (32) *I have in a sultry way kissed him

    It is possible, then, that -lymanner adverbs and the whole class of PP

    adverbials are differentiated in their syntactic position. This discussionsuggests that the manner adverbs under consideration are syntactically

    positioned close to the surface position of the verb, while locatives are more

    distant.

    There is evidence, though, that locatives at least are actually lower down

    than the surface position of the object. Note that in (33), the quantifier in

    object position can bind the pronoun in the locative, suggesting it c-

    commands it (see also Pesetsky 1995, p161):

    (33) Maire tortured every rabbitiin itsihutch

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    6/25

    6

    A theory which assumes that locatives are right adjoined will have to

    deal with this data by assuming that the object raises to a position higher upthan VP (perhaps [Spec, AgrO]), and that the verb raises higher still.

    This set of data forms the basic desiderata of a theory of the positions of

    these adverbials. The conservative position is that manner and locative are

    both adjoined low to VP, and that the object and subject raise to positions

    outside VP:

    (34)

    TP

    DP T

    T FP

    F AgrOP

    DP i AgrO

    Vk

    AgrO VP

    VP LOC

    VP MAN

    tk ti

    However, this approach does not explain why we find the extraction

    contrast discussed above, since both Locatives and manners are generated

    adjoined to VP. In addition, this approach cannot tie the putative differentialpositions of manners and locatives down to their syntactic licensing, since

    nothing is said here about licensing, and indeed the most recent proponent of

    this kind of structure puts the fact that manners are closer to the surface

    position of the verb than locatives down to purely semantic factors (see, e.g.

    Ernst 2001).

    3. Recent accounts

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    7/25

    7

    Less conservative positions have also been adopted recently. The

    structure in (34) of course involves right adjunction, an option which UG

    has been argued to lack (see, notably, Kayne 1994). Right adjunction isbarred under Kayne's assumptions because adjunction is barred, and because

    rightward merge to a head is barred unless the merged element is a

    complement. These constraints follow from a particular view of the way

    that PF linearisation is read off syntactic structures.

    Cinque (1999) develops some of Kayne's ideas within the context of a

    theory of adverbs, and essentially argues that adverbials are to be found in

    left specifier positions of various functional heads. He points out two ways

    of dealing with manner and locative adverbials: either they are generated as

    specifiers of light verb shells within the VP, or they are generated as

    complements of these heads, with the VP in their specifier (an idea he

    attributes to Nilsen 1998).Nilsen's proposal is similar to ideas first proposed within the Generative

    Semantics tradition by Geis (1970), and it is an idea he extends to all

    circumstantial adverbials. The core notion is that light v heads are involved

    which mediate a predication relation between the VP, which has already been

    constructed, and the adverbial. So for a sentence like (35) the VP "smoked

    banana-skins" is in the specifier of a light verb whose complement is "in the

    park" and whose semantic function is to locate the event denoted by the VP

    at the appropriate spatial point:

    (35) Johnny smoked banana-skins in the park

    (36)

    vP

    VP v

    v PP

    smoked banana-skins

    In the park

    This is, in many ways, an attractive idea, and leads to a fairly clean

    semantics for these adverbials. However, note that it still says nothing about

    the relative hierarchical ordering of manners and locatives. In addition, there

    is no evidence for the projection of these extra v heads: they have no

    phonological content, and their semantics is either vacuous, or uniformly

    predicational. If the latter, then it is impossible even to state the ordering of

    manners and locatives, since to do so would require the v heads to have

    different semantics from each other, or different syntactic properties.

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    8/25

    8

    A further problem here is the binding data noted above. Nilsens

    approach would assign a sentence like (33) the structure indicated:

    (37) Maire [tortured every rabbit] [in its hutch]

    In such a structure, no c-command relation holds between the quantifier

    phrase and the pronoun it binds.

    Cinque's own proposal suffers from the same problem. He suggests that

    this type of adjunct is generated in the specifier of a light verb shell above

    VP:

    (38)

    vP

    PP v

    In the park v vP

    AdvP v

    v VP

    quickly

    smoked banana skins

    Note that in order to achieve the correct word order, we need to posit

    further functional structure. The idea is that VP will raise into the specifier

    of a functional head (not shown above) immediately above the head that

    licenses the manner adverb, and then the projection of this head will raise

    into the specifier of a further head above the locative, giving rise to a type of

    'leapfrogging' movement. Again, we reject this proposal on the grounds that

    it requires more functional structure than is motivated by the phonology or

    semantics of the constructions, and moreover it suffers from the sameempirical problem we saw with Nilsens account, since the object will not c-

    command a bound pronoun in the locative.

    4. An alternative

    The three systems briefly discussed above, of course, were not primarily

    concerned with dealing with the data we outlined in section 1. However, we

    think that proper attention to this data actually allows us to motivate the

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    9/25

    9

    correct functional structure required to give an answer to questions of how

    these adverbials are licensed.

    Following much recent work, we adopt an articulated structure for the

    verb phrase, consisting of the lexical VP, surmounted by a number of

    functional heads which encode particular semantic relations. In particular, we

    adopt the idea defended by Travis (1991, 2000) and others (Borer,

    forthcoming, Ritter and Rosen 1998) that aktionsart (lexical aspect) is

    marked syntactically by an aspectual functional head which takes the verb

    itself as its sister. This head is itself the complement of the Agent

    introducing head little v (Kratzer 1995, Hale and Keyser 1993, Chomsky

    1995). This gives us the following structure:

    (39)vP

    v AspP

    Asp VP

    Johnson (1991), Lasnik (199), and others, argue, on the basis of diverse

    phenomena such as gapping and the interaction between particles and

    double object constructions, that the main verb in English raises out of the

    verb phrase proper to some higher head. The precise semantic nature of this

    head is not relevant for the discussion here, but we take it to encode sentence

    aspect, as opposed to the lexical aspect mentioned above. For convenience

    we will label the higher head Asp1 and the lower (the vp-internal one)

    Asp2.

    The full structure we adopt for the English verb phrase, then, is:

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    10/25

    10

    (40) Asp1P

    Asp1 vP

    DP v

    V

    v Asp2P

    Asp2 VP

    tv

    V DP

    tv

    tv

    The DP in [Spec, vP] is interpreted (at least for some varieties of v, see

    Kratzer 1995) as the agent, and the verb raises through Asp2, and v to Asp1

    in English.

    Our analysis of the adverbial facts will be couched in terms of the

    framework outlined in Chomsky (2000). In this framework, functional

    heads are assumed to bear features which set up dependencies with

    formatives that the head c-commands. These dependencies are formed when

    the functional head concerned is specified with uninterpretable features.

    These features are termed the probe. A probe essentially seeks matching

    features within its c-command domain (these matching features are the goal).

    The relationship between probe and goal is constrained by locality.

    Essentially, the relevant kinds of structures are like those in (41):

    (41) [ H{probe} [ ..... XP{goal} ... ]] (probe=goal)

    The relation between the head H specified with the probe, and the

    formative specified with the goal, we will call the H-associate relation,

    extending Chomskys terminology. The formation of an H-associate

    relationship results in the deletion of the uninterpretable features involved in

    the relationship. Since it is the probe that is uninterpretable, the probe

    deletes.

    In addition to probes, heads may also be specified with EPP features.

    These features are selectional (i.e. involve category information) and are also

    uninterpretable. An EPP feature is satisfied when a category of the

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    11/25

    11

    appropriate featural specification is merged with the head bearing the

    feature. The XP that is merged can be either the goal itself, or some other

    phrase (for example, an expletive):

    (42) [ XP H{probe, EPP} [ ..... XP{goal} ... ]] (probe=goal)

    The deletion of EPP feature is, in general, parameterised, so that some

    languages allow multiple subject constructions, as discussed in section 4.10.3

    of Chomsky (1995). The system outlined there can be thought of as a set of

    conditions on the deletion of the EPP feature: a language does not have EPP

    (VSO languages); has EPP but merge into [Spec, HP] causes EPP to delete

    (SVO without Multiple Subject Constructions (MSCs)); has EPP but allows

    one element to merge without deleting EPP (SVO with MSCs); or allows

    arbitrarily many merges without deleting EPP (polysynthetic languages). Wewill adopt Chomskys idea that the EPP feature must be satisfied, and that

    when an H-associate relation is set up, the XP (usually determined by the

    goal of Hs probes) will be forced to raise to merge with the projection of H,

    satisfying EPP. In this situation, EPP deletes.

    We will, however, extend this picture, adapting Chomskys idea that

    deletion of EPP is an option that UG allows variation for. The core

    extension is the assumption that, when EPP is satisfied by an element which

    hasn't induced the H-associate relation, the EPP feature does not have to

    delete immediately4. Within a phasebased model (Chomsky 2001), we

    may suppose with Pesetsky and Torrego (2001) that deletion of features

    takes place at the phase level. Thus, for what concerns us here, an EPP

    feature that has not deleted immediately remains active until the current

    phase is completed at which point the feature is deleted, perhaps as part of

    the TRANSFER operation of Chomsky (2002).

    In essence, then, a single EPP feature may trigger Merge of an arbitrary

    number of non-Agreeing XPs. If no Agreeing XP is merged at all, the EPP

    feature is deleted at phase-level. If an Agreeing XP is Merged, then EPP is

    deleted immediately.

    It follows that we allow an arbitrary number of non-agreeing XPs to

    merge with H, in the general case:

    (43) [ XP YP ... ZP H{probe, EPP} [ ..... XP{goal} ... ]]

    (probe=goal)

    In (43) EPP will only be forced to delete when XP is merged with the

    projection of H, since only XP is determined by the goal of Hs probe.

    This system gives rise to a potentially infinite number of adverbials in

    inner specifier positions, constrained by only processing considerations, and

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    12/25

    12

    the particular specification of the EPP feature (i.e. what it is a selectional

    feature for).

    Before seeing how this picture pans out in detail, we should state howthe semantic interpretation of these structures is governed. Adapting

    proposals by Borer (1998a, b) among others (McClure 1995, van Hout

    1996), let us assume that a telic interpretation of Asp arises when Asp's

    probe finds a matching goal. Put another way, only telic specifications of

    Asp have an uninterpretable probe which will match features of the object.

    The features that are relevant, in this case, are features governing the

    quantization of the object (following Verkuyl 1993). We will call this feature

    [Quant] and assume that it is interpretable on DP, but not on telic Asp. The

    specification of telic asp is then Asp[uQuant, EPP], where the u prefix

    signals that the feature is uninterpretble (following the notation of Pesetsky

    and Torrego 2001). This allows us to capture the well-known fact thatquantized objects give rise to telic readings of certain predicates:

    (44) We built that house (telic)

    (45) We built houses (atelic)

    Having outlined the basic system of assumptions we adopt, the

    particular cases to be considered are as follows: when Asp's probe finds a

    matching goal in the VP but no locative phrase has been constructed from the

    numeration, the Asp-associate relation is established, a telic interpretation

    results, and the object raises to [Spec, Asp] to satisfy the EPP feature of

    Asp as in (46) (irrelevant details omitted):

    (46)

    vP

    DP v

    v Asp2P

    DP Asp2

    Asp2 VP

    V DP

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    13/25

    13

    Since an Asp-associate relation has been established, the EPP feature of

    Asp deletes and no further Merge into [Spec, AspP] is possible. However, ifa locative phrase is accessible at a point in the derivation before DP raises, it

    may satisfy the EPP feature of Asp. By accessible at this point of the

    derivation, we simply mean previously constructed in the derivation, or

    accessible as a simple lexical item from the numeration. It may be the case

    that these locatives have been adjoined to VP, and move to an inner [Spec,

    AspP], but this is not necessary, and in fact might be ruled out, since

    nothing will force them to Merge with VP. The minimal solution is that they

    are simply present in the workspace of the derivation.

    Note that the EPP feature of Asp must be of the correct selectional type

    to allow a locative to Merge, and we will assume that locative PPs and

    quantized DP objects both are specified with interpretable (and thereforeselectable) features which relate an event and an individual in terms of spatial

    measure: a locative measures out the physical extent over which an event is

    delimited, while a quantized DP measures out the physical extent of the

    result of the event. Locatives, then, are specified as [Quant]. Of course, the

    idea that certain predicates can select for particular semantic properties like

    this, and that such selection results in particular prepositions, or particular

    case forms, is a traditional one. The preposition within the locative

    establishes a probe-goal relationship with the DP in terms of its case

    requirements. Asps [uQuant] feature, on the other hand, cannot match with

    [Quant] on DP because [Quant] on the locative P intervenes.

    Our proposal leads to the possibility that a number of locatives may be

    merged with Asp. However, once the object has been merged, then the EPP

    feature deletes, and no further merge of locatives is possible, since merge of

    locatives is only made possible by the presence of an EPP feature. This

    gives the basic structure:

    (47) [Object Locative Locative Asp t]

    This predicts that the object will c-command the locative, as we noted in

    section 2 with respect to example (33) repeated here as (48):

    (48) Maire tortured every rabbit in its hutch.

    Consider now a case where there is no object within the verb phrase, but

    where a locative has been constructed. The locative will merge with Asp,

    satisfying its EPP feature, but no telic interpretation will result, because no

    probe-goal relation has been established. Asp, in this case, does not bear

    [uQuant] and is therefore not (obligatorily) telic. This is precisely what is

    behind the locative preposition drop phenomenon repeated below:

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    14/25

    14

    (49) a. They climbed up the mountain.

    b. They climbed the mountain.

    (49a) converges because the EPP feature of Asp is satisfied by the

    locative. The locative preposition enters into a P-associate relation with the

    f-features of the DP the mountain resulting in it being essentially Case-

    licensed (more on which, see below). Asp itself cannot enter into an Asp-

    associate with this DP, since this would give rise to a locality violation (the

    preposition bearing [Quant] is closer). This, in turn, means that Asp cannot

    be specified as telic since only telic asp has a quantization probe, giving rise

    to the appropriate interpretation.

    If there is no object, we have a violation of EPP:

    (50) They climbed

    (50) can only be construed as unspecified object drop, the unspecified

    object satisfying EPP of Asp, but not giving rise to quantization effects;

    hence (50) may be interpreted as atelic.

    (49b), on the other hand, does involve an Asp-associate relation with the

    object DP, EPP is satisfied and the derivation converges.

    Let us now turn to the alternation in Hebrew, discussed by Borer. Recall

    that a lexically telic verb allowed inversion, but that an atelic one did not,

    unless it occurred with a locative clitic:

    (51) a. Parcu mehuma (telic)

    Erupted riot

    "A riot erupted"

    b. *avad ganan (atelic)

    Worked gardener

    "A gardener worked"

    c. avad sam/kan/ecli ganan

    Worked here/there/chez-me gardener

    "A gardener worked here/there/at my house"

    On our account, a verb likeparcu "erupt" lexically selects an Asp withactive probes (that is, it bears an uninterpretable quantization feature) while

    a verb like 'avad "work" does not (this distinction is in addition to the

    different Merge positions of the single argument in each case). In (51a), the

    probe of Asp establishes the Asp-associate relation with the object (which

    then raises to satisfy the EPP feature of Asp). The object bears a case

    feature which needs to be checked. Following Chomsky (2001), we assume

    that case checking (which we can implement as deletion of a uCase feature)

    is parasitic on H-associate relationships established in terms of j-features

    only (see the discussion of the case licensing properties of little v below).

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    15/25

    15

    We then have the following schematic derivation:

    (52) Asp[uQuant, EPP] DP[f, Quant, uCase] =>(53) Asp[uQuant, EPP] DP[f, Quant, uCase] =>

    (54) DP[f, Quant, uCase] Asp[uQuant, EPP] t

    In (51), no Asp-associate relation is established because the subject is

    Merged in the specifier of little v, and is therefore not c-commanded by Asp,

    and the single argument ganan must instead be checked by T's f probes,

    which also case license it (note, since this verb is agentive, the DP argument

    is merged in [Spec, vP] and is therefore too high to be a goal for Asps

    probes):

    (55) T[uf, EPP] DP[f, Quant, uCase] =>

    (56) T[uf, EPP] DP[f, Quant, uCase] =>

    (57) DP[f, Quant, uCase] T[uf, EPP] t

    If this relationship is established, ganan must raise to satisfy the EPP

    features of T.

    The next question is how the EPP features of T are satisfied in (51a)? In

    this case some null element must fill the specifier of TP, and we follow Pinto

    (199*), who convincingly argues on the basis of Italian, that verbs of this

    aspectual class can always select a covert locative. For us, this locative is

    merged to satisfy the EPP feature of telic Asp. From this position it raisesto the specifier of T. The appearance of inversion here, then, arises because

    of a null locative in subject position, rather than the subject itself.

    Turning to (51c), this example is parallel to (51a), except that, again

    following Pintos work, the null locative is not available for atelic predicates.

    However, it is of course possible to select an overt locative in the

    numeration, and in this case this can satisfy the EPP feature of T, so that the

    subject remains in situ, giving rise to the apparent inversion.

    This system captures the core of the semantic and syntactic

    dependencies between locatives, objects and aspectual specification,

    although clearly there is more to be said. The system also automatically

    derives the apparent iterability of locatives, as well as the fact that objects

    must c-command locatives. In addition, the structures proposed involve only

    heads that are interpretable at the interfaces, rather than appealing to

    semantically empty functional structure.

    If we assume that T behaves in a similar fashion to Asp, then the system

    we have outlined leaves open the possibility that locatives may satisfy an

    EPP feature of T, after establishment of a T-associate relation with the single

    argument of an intransitive, as in the Hebrew example discussed above. In

    such cases, as in the cases above, the locative and subject are in the same

    minimal domain and therefore equidistant to the probes of T. Under this

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    16/25

    16

    scenario, T establishes an associate relation with both, but only one raises to

    satisfy (and delete) the EPP feature of T. The familiar case of Locative

    Inversion (Bresnan 1994) may exemplify exactly this situation:

    (58) a. In the garden sat a gnome

    b. A gnome sat in the garden

    Note that the probe of T here would have to be person features only,

    since agreement is not triggered by conjoined inverted locatives. Of course,

    factors such as focus play an important role in Locative Inversion

    constructions, as does the thematic structure of the predicate (something we

    take as being at least partially reducible to aspectual considerations).

    However, the suggestion seems promising.

    Let us turn now to manner adverbials. As noted in section 2, -ly manneradverbs appear closer to the verb than locatives do, although there is no clear

    evidence that -ly manners c-command locatives or vice versa, and it appears

    that prepositional manners and locatives appear in the same position.

    Recall that Asp induces an Asp-associate relationship with the object.

    The goal in the object we took to be features associated with semantic

    quantization. We did not assume that the goal of Asps probes was related

    to Case, or to f-features. In our system, it is v that establishes a v-associate

    relationship with these features of the object. The licensing of the object

    takes place, then, via at least two different featural relationships between

    functional heads and different goals within the object. Asp probes forquantization, while v probes for f-features.

    To maintain maximal parallelism between the components of the

    extended verbal projection, v also has an EPP feature, which is satisfied and

    deleted by an XP determined by the goal of vs probes in this case the

    object. This is the standard assumption within Minimalist approaches to

    clause structure (see the extended discussion in Chomsky 1995, 2000, and

    Lasnik 1995s argument that movement to [Spec, AgrP] is always driven by

    EPP considerations). The object will then raise to Spec, vP, which we

    assume is its surface position in English (Johnson 1990, Koizumi 1993, and

    the papers collected in Lasnik 1998), with the verb raising outside the verb

    phrase, as discussed in the introduction to this section.

    This gives us the following structure:

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    17/25

    17

    (59)Asp1P

    Asp1 vP

    vP

    SUBJ

    V OBJ v

    v Asp2P

    tv LOC Asp

    Asp2 VP

    V

    tOBJ tv

    tv

    In the same way that locatives satisfy the EPP feature of Asp without

    causing it to delete, -ly manner adverbs will satisfy the EPP feature of v,

    again without inducing deletion. This means that manner adverbials are

    inserted into the structure as inner specifiers of v.

    What is the actual structure of Manner adverbials themselves, and whyare they licensed by the EPP feature of v? We tentatively suggest that

    manners are actually nominalisations of copies of the verb in the numeration,

    which are modified by adjectival modifiers. So a manner like slowly in the

    phrase slowly ran is essentially derived by copying the root ran in the

    derivation, modifying it with an adjectival predicate slow and then

    projecting a nominal functional structure above the new composite, so that

    the whole phrase is morphologically interpreted as slowly. This process

    can be seen overtly at work in many languages (see, for example, the

    discussion of manner modification in Modern Hebrew, in Glinert 1995), and

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    18/25

    18

    makes sense of the fact that manners relativise to their predicate, so that

    slowly running might involve speedier movement than quickly walking.

    Since the resulting phrase is nominal, it may satisfy the nominal EPP featureof v.

    (60)

    Asp1P

    Asp1 vP

    vP

    SUBJ vP

    V OBJ vMAN

    v Asp2P

    tv LOC Asp

    Asp2 VP

    V tOBJ tv

    tv

    Inspecting the structure above, we predict a word order for English

    which is V Obj MAN LOC. This is of course the correct prediction. Notethat in this structure the manner adverb c-commands the locative, something

    that we argued was at least a possibility in section 2.

    We also immediately predict that it is possible to iterate manner

    adverbials, as can be seen in the following examples:

    (61) They played loudly badly (from Ernst 2001).

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    19/25

    19

    This is, of course, because of the way that EPP features of functional

    heads can be satisfied, but not deleted, by elements that do not establish H-

    associate relations with the functional heads probe.How do we account for the correlations noted in section 1, between the

    case licensing potential of the verb and its ability to occur with manner

    adverbs? Recall that the probe goal relationship set up by v is similar to the

    traditional case relationship (although technically it will involve

    uninterpretable phi-features, rather than case features). Clearly, a version of

    v which is not endowed with the appropriate probe will not license a case

    marked object (and such verbs will not have an Agent in their specifiers

    either, following the usual implementation of Burzios generalization in this

    framework Chomsky 1995). It follows that verbs like resemble etc. will

    not check structural accusative case of their object, thus accounting for their

    anomalous behaviour in of-insertion environments. There are a number ofways to implement this, but the simplest is just to assume that these verbs

    do not have a v embedded in their structure at all. Given that these

    predicates have an impoverished structure, specifically lacking in an EPP

    feature, they will not be able to license manner adverbials, explaining the old

    observation that predicates which do not assign structural accusative, do not

    take manner adverbials.

    (62) *John resembled Sue slowly.

    (63) *John had flu worriedly.

    (64) *The slave cost 600 denarii wholeheartedly

    If passivisation is simply an operation on v, say deleting the probe of v,

    then we also explain why these predicates do not passivise:

    (65) a. *Sue was resembled by John

    b. *Flu was had by Maria

    c. *50 were cost byRhyme and Reason.

    One final point to note is that, as mentioned in section 1, the kind of

    system we adopt assumes that agentive unergatives do actually contain a

    little v (cf. Hale and Keyser (1993), who argue that such verbs are actuallytransitive where the object has incorporated into the verb). It follows, then,

    that it is possible to have a manner adverb in the absence of an affected

    object:

    (66) John walked jerkily to the cliff edge.

    Following Hale and Keyser (1993), we assume that such unergatives do

    contain a little v, which is the licenser of the manner adverb.

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    20/25

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    21/25

    21

    NOTES

    1A much earlier version of this paper appears as Adger and Tsoulas (2000)..

    Many thanks to two anonymous reviewers and to the volume editors for

    helpful comments.2Note that in cases like (i), only an extent interpretation is available for the

    adverb:

    (i) John had flu badly.

    3Note that there are verbs which disallow of-insertion but which are fine

    with adverbs. These seem to be copular verbs:

    (i) *John's becoming of a fool/John's being of a fool

    (ii) John became a Buddhist willingly/John was a doctor wholeheartedly

    Presumably the lack of of-insertion arises because these verbs take

    predicates as their complements, rather than true objects.4This is also reminiscent of the theory of adjunction developed in Saito and

    Fukui (1998) where they argue that multiple adjunction at the X level ispossible until a specifier which agrees with the head is merged. In their

    terms, an agreeing specifier closes off the projection.

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    22/25

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    23/25

    References

    Adger, David & George Tsoulas. (2000) Aspect and lower VP adverbials. InArtemis Alexiadou & P. Svenonius (eds) Adverbs and Adjunction, pp.

    1-18. Linguistics in Potsdam 6. Potsdam.

    Andrews, Avery. (1982) A note on the constituent structure of adverbials

    and auxiliaries. Linguistic Inquiry 13/3, 313 317.

    Belletti , Adriana. & Luigi Rizzi (1988) Psych-verbs and Theta Theory.

    Natural Language and Linguistic Theory 6/2, 291 352.

    Borer, Hagit. (1998a) Deriving passive without theta roles. In: Steven

    Lapointe et al. (eds.) Morphology and its relation to phonology and

    syntax, CSLI: Stanford.

    -- (1998b) Licensing Aspectual Nodes Paper presented at the Afro-Asiatic

    Linguistics Conference, SOAS, London.--, forthcoming, The ExoSkeletal Trilogy, OUP.

    Bowers, John. (1993) The syntax of predication. Linguistic Inquiry

    24/4:591-656

    Bresnan, Joan. (1994) Locative inversion and the architecture of universal

    grammar. Language, 70/1: 72 131.

    Chomsky, Noam. (1965) Aspects of the Theory of Syntax. Cambridge,

    Mass. MIT Press

    -- (1995) The Minimalist Program Cambridge, Mass. MIT Press

    -- (2000) Minimalist Inquiries. In Roger Martin et al. (eds) Step by Step:

    Essays on Minimalist Syntax in Honor of Howard Lasnik, pp.89-156.

    Cambridge, Mass. MIT Press.

    -- (2001) Derivation by Phase. In Michael Kenstowitz (ed): Ken Hale: A

    Life in Language. pp. 1-52. Cambridge, Mass. MIT Press.

    -- (2002) Beyond Explanatory Adequacy. Ms MIT.

    Cinque, Guglielmo. (1999) Adverbs and Functional Heads. Oxford, Oxford

    University Press

    Costa, Joao. (1996) Adverb positioning and V-movement in English: some

    more evidence. Studia Linguistica, 50/1, 22 34.

    Ernst, Thomas. (1984) Towards an integrated theory of adverb position in

    English, IULC, Bloomingdale.

    -- (2001) The syntax of adjuncts. Cambridge, Cambridge University Press.Geis, John (1970) Lexical Insertion of locative and time prepositions,

    Proceedings of the Chicago Linguistic Society, Vol 8.

    Glinert, Lewis (1995) The grammar of Modern Hebrew,Cambridge, CUP.

    Haider, Hubert, 1997, Extraposition, in Dorothee Beerman, David LeBlanc

    and Henk van Riemsdijk (eds) Rightward Movement, Amsterdam: John

    Benjamins. 115-151.

    Hale, Ken. & Samuel. J. Keyser (1993) On argument structure and the lexical

    expression of syntactic relations. In: Ken Hale & Samuel J Keyser (eds)

    The view from Building 20; Cambridge, MIT Press.

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    24/25

  • 8/14/2019 Adger and Tsoulas- Circumstantial Adverbs and Aspect

    25/25

    Verkuyl, Henk. (1993), A theory of aspectuality: the interaction of temporal

    and atemporal structure. Cambridge, CUP.

    Zubizarreta, Maria-Luisa (1998) Prosody, Focus and Word Order, MITPress