Result at Ive

download Result at Ive

of 21

Transcript of Result at Ive

  • 7/29/2019 Result at Ive

    1/21

    1

    ADJECTIVE SELECTION IN RESULTATIVECONSTRUCTIONS

    Cristiano Broccias

    0. INTRODUCTION

    The apparently rampant idiosyncrasy in the use of adjectives in resultative

    constructions (e.g.John hammered the metal flatvs. *John hammered the metal

    long, see sections 1 and 2) has often led researchers to despair of findingconstraints (or generalizations) capturing their occurrence. Quite recently,

    however, Wechsler (2001) has developed a formal model which predicts many

    cases of adjective selection, as will be shown in section 3. In this paper,however, I will show that Wechslers analysis, although basically on the righttrack, is untenable if taken as a formal system (see section 4). In its place, I will

    propose that adjective selection can be captured by taking into account, among

    many other factors, the distinction between gestalt and part-whole properties

    (see section 5).

    1. RESULTATIVE CONSTRUCTIONS

    Resultative constructions are sentences such as (1):

    (1) John hammered the metal flat.

    The adjectiveflat in (1) describes the state achieved by the metal as a result of

    the action of hammering performed by the subject referent (i.e. John). Hence,flatis referred to as a resultative (adjective) phrase and the construction in which

    it appears as a resultative construction (see Levin 1993: 101). In other words, (1)

    codes two subevents: the subevent of Johns hammering the metal and the

    subevent of the metals becoming flat. Such two subevents are related causally:the former subevent causes the latter subevent (i.e. the change of state event).

    The situation is informally represented in (2), where the thick arrow visualises

    the causal relation:

  • 7/29/2019 Result at Ive

    2/21

    2

    (2) SUBEVENT1: John hammered the metal

    SUBEVENT2: the metal became flat

    It must be pointed out that the constructional object need not be a possible

    object of the verb when this is used in isolation, that is independently of theresultative construction. Let us consider the data in (3):

    (3) a. Sally wiped the table clean.

    a. Sally wiped the table.

    b. Sally laughed herself silly.

    b. *Sally laughed herself.

    The constructional object the table in (3a) is also a possible object of the

    verb wipe when this verb is used in isolation, i.e. independently of a resultativeconstruction, compare (3a) with (3a). In keeping with current linguistic usage

    (see Haegeman 1994), I refer to this type of object as a subcategorised object,which means that it is selected by the constructional verb. The constructional

    object herself in (3b), on the other hand, is not a possible object of the verblaugh when such a verb is used in isolation, as is shown in (3b). Herselfin (3b)is therefore called unsubcategorised object since it is not selected (or

    subcategorised) by the constructional verb. Finally, since both the table andherself refer to entities that underwent some change, these can be described

    through the semantic roles ofpatient or theme1. The latter term reflects theintuition that the affected entity in the resultative construction (i.e. the object in(3a) and (3b)) can be viewed metaphorically as traversing a path, whose

    endpoint is signalled by the adjective (see Broccias 2003 for more details).

    2. IDIOSYNCRASY

    Corpus-driven analyses of resultative adjectives (i.e. adjectives occurring inresultative constructions), such as the one carried out by Boas (2000) using the

    British National Corpus (BNC), show that a very restricted range of adjectives isused. Table 1 reproduces the resultative adjectives found in the BNC by Boas

    (2000)2

    :

    1The labels for thematic roles are usually capitalised. I will not follow such a convention in

    this paper.2

    It must be pointed out that the resultative adjectives found by Boas (2000) are not

    necessarily all the ones occurring in the BNC. The reader is referred to his work for details on

  • 7/29/2019 Result at Ive

    3/21

    3

    We observe that the first five adjectives alone (i.e. dead, apart, open, shut,

    clean) account for 82% of all adjectival occurrences and that, phonologically,the adjectives in Table 1 are either mono- or bi-syllabic. The only exception is

    unconscious, which is however formed from the bi-syllabic adjective conscious

    by adding the negative prefix un-. Finally, only one adjective, crooked, isdeverbal.

    Not only does the range of adjectives found in resultative constructions seem

    to be very limited statistically, but native speakers judgements reveal that

    potentially plausible adjectives do not actually occur in resultative constructions

    (at least if uttered out of the blue). Consider the following minimal pairs:

    (4) a. John hammered the metal flat. (= (1))

    b. *John hammered the metal {long/triangular}.

    (5) a. John laughed himself to death.

    b. *John laughed himself dead.

    Table 1 Resultative adjectives found in the BNC (from Wechsler 2001,based on Boas 2000)

    adjective occurences adjective occurences

    dead 431 thin 15

    apart 407 sick 13open 395 red 11

    shut 207 stupid 10

    clean 102 hoarse 9

    dry 77 smooth 5

    awake 41 empty 4

    full 35 crooked 2

    flat 34 deaf 2

    unconscious 30 sober 1

    black 25 soft 1

    silly 20 sore 1

    Although a scenario in which John hammered the metal thus causing it to

    become long or triangular can easily be activated on the part of theconceptualiser, (4b) is judged ungrammatical. (5) shows that in some cases a

    the methodology employed.

  • 7/29/2019 Result at Ive

    4/21

    4

    morphologically related noun (used as a prepositional complement) is preferredto the adjective.

    On the basis of evidence similar to that provided in (4) and (5), researchers

    have concluded that adjectival selection in resultative constructions is highly

    idiosyncratic (see Verspoor 1997 among others). Only recently has Wechsler(2001) developed a model which can (partially) capture adjectival distribution inresultative constructions. In what follows, I will detail Wechslers models

    predictions and show that they are not always borne out by actually occurring

    examples. Further, Wechslers model has little to say about unsubcategorised

    object cases (see (3b) above), which, I will contend, are sensitive to the samegeneralisations as subcategorised object examples.

    3. WECHSLERS (2001) MODEL

    In this section I will illustrate the predictions made by Wechslers (2001)

    formal model but will not discuss its theoretical underpinnings because this goeswell beyond the scope of the present paper (the interested reader is referred to

    Wechslers own paper and Broccias 2003 for detailed discussion). Still, it iscrucial to point out that Wechslers analysis is based on the observation that

    resultative constructions are always telic or, to put it differently, they always

    involve a telic path. The sentence in (6)

    (6) *John hammered the metal flat for an hour.

    is impossible because the temporal adjunctfor an hourimposes a repetitive (i.e.

    atelic) construal for the event coded by the resultative construction (i.e. the

    metals becoming flat). But such an event, under ordinary circumstances, cannottake place more than once.

    Wechslers model predicts two ways in which the verbal (or causing)event can interact with the telic path (which is expressed by the resultative

    phrase):

    Case 1. If the theme argument is subcategorised (e.g. the theme is an object of

    the verb in isolation)3, then homomorphism and coextension between the event

    3The theme argument can also be shared when it is the subject of an intransitive verb (as in

    The river froze solid). This case is not discussed in the present paper, which only focuses on

    transitive resultative constructions. Nevertheless, the conclusions arrived at here also apply to

    intransitive examples (see Broccias 2003 for discussion).

  • 7/29/2019 Result at Ive

    5/21

    5

    and the path is required (i.e. the verbal, or causing, event, SUBEVENT1, and thechange event, SUBEVENT2, unfold together).

    Case 2. If the theme argument is not subcategorised, then homomorphism and

    coextension between the event and the path is not required (i.e. the verbal, orcausing, event and the change event need not unfold together).

    Homomorphism means that parts of the verbal event must correspond to parts of

    the path and vice versa. For example, in (3a) above, Sally wiped the table clean,

    parts of the event of wiping correspond to parts of the event of the tablesbecoming clean, i.e. the path metonymically signalled by the adjective clean

    (which denotes its final point). The term coextension in the definitions abovemakes it clear that the verbal event must begin when the affected theme is at the

    start of the change path and end when the affected theme reaches the end of thechange path. In other words, homomorphism between the telic event and the

    path obtains at the same arbitrary point in time along a(n initially and finally)bounded time arrow. For the sake of simplicity, I will refer to homomorphism

    and coextension simply as homomorphism. I will now turn to a detaileddiscussion ofCase 1 and Case 2 examples.

    3.1. Case 1:Subcategorised arguments

    Case 1 involves sentences such as (1), John hammered the metal flat, and(3a), Sally wiped the table clean, where the constructional object is also a

    possible object for the verb used in isolation. We can distinguish three types ofCase 1 sentences according to the temporal profile of the verbal event (i.e.

    durative vs. punctual event) and the lexical category of the head in theresultative phrase (i.e. adjective vs. preposition). The first type, Type I, is

    illustrated in (7) below; it combines a durative verbal event with an adjectivephrase.

    (7) a. Sally wiped the table clean. (= (3a))

    b. *The rabbits had apparently been battered dead.

    Since wipe in (7a) is a durative verb, for homomorphism to obtain between the

    verbal event and the path hinted at by the resultative adjective (i.e. the pathending up in the clean state), a suitable, i.e. durative, adjective must beselected. In more detail, Wechsler contends that the adjective in question must

    be gradable (cf. cleaner) and closed-scale (cf. completely clean). Informally, theformer requirement correlates with path extension for the change event hinted

  • 7/29/2019 Result at Ive

    6/21

    6

    at by the adjective, whereas the latter correlates with boundedness of thechange event. A schematic visual representation for (7a) is offered in Figure 1,which will also be appropriate for Type IIIbelow.

    Figure 1 Schematic representation of Wechslers (2001) Type Iand TypeIIIresultative constructions

    The horizontal lines in Figure 1 stand for temporal (or path) extension and thevertical lines at either end of the lines indicate that both the verbal event (the

    upper horizontal line) and the resultative path (the bottom line) are bounded.

    The dashed vertical line connecting the two horizontal lines, as well as the factthat the latter have been represented one exactly under the other and as having

    the same length, visualises homomorphism. Going back to (7), we are now in aposition to explain the unacceptability of (7b),*The rabbits had apparently been

    battered dead. The verb denotes a durative event, but the adjective does not.Dead is a non-gradable adjective (cf. *more dead). In other words, the path

    coded by deadhas no extension and would correspond to a point (rather than a

    segment) in a diagrammatic representation along the lines of Figure 1 (seeFigure 2 below).

    Type IIinvolves cases where the verbal event is punctual and the resultative

    phrase is an adjective phrase. For homomorphism between the verbal event and

    the resultative event to obtain, the adjective must be punctual as well, that isnon-gradable, as is illustrated in (8) below:

    (8) Pat shot the miller dead.

    A schematic visual representation ofType IIsentences is given in Figure 2.

    gradable adj. / PP

    durative verb

    homomorphism

  • 7/29/2019 Result at Ive

    7/21

    7

    Figure 2 Schematic representation of Wechslers (2001) Type IIresultative constructions

    The punctual nature of both verb and adjective has been depicted in Figure 2

    through a cross. The dashed vertical line, as in Figure 1, indicates

    homomorphism.Type III concerns those cases where the resultative state is encoded in a

    prepositional phrase (PP). According to Wechsler (2001), since prepositional

    phrases always code durative events, the verbal event must also be interpreted as

    such. This explains why the verb shoot, which occurred with a punctual

    interpretation in (8) above, has a durative reading in (9a) in similar fashion tothe by-default durative verb batterin (9b).

    (9) a. Pat shot the miller to death.

    b. Pat battered the rabbit to death.

    Wechsler (2001) claims that his model explains the unacceptability of the

    examples in (10).

    (10) a. *He hammered the metal safe. (no homomorphism)

    b. *He hammered the metal beautiful. (no homomorphism)

    c *He hammered the metal tubular. (cf. ??very tubular)

    d. *He hammered the metal triangular. (cf. ??very triangular)

    e. *He hammered the metal wide. (cf. ??extremely wide)

    f. *He hammered the metal long. (cf. ??extremely long)

    punctual verb

    punctual adj.

  • 7/29/2019 Result at Ive

    8/21

    8

    Such problematic examples involve either lack of homomorphism, see (10a)

    and (10b), or an unsuitable adjective (i.e. a non-gradable or non-closed scaleadjective, as indicated for (10c)-(10f) by the deviant expressions in brackets).

    We can already note at this juncture that Wechslers claim concerning the lack

    of homomorphism for (10a) at least is intuitively dubious. To be sure, thehammering event is coextensive with the metals ending up in a safe position.

    3.2. Case 2:Unsubcategorised arguments

    Case 2 concerns those cases where the constructional object is not a possible

    object for the verb used in isolation, that is the constructional object is anunsubcategorised object. Sentences such as

    (11) a. Sally laughed herself silly. (= (3b))

    b. Sally danced her legs stiff.

    do not imply homomorphism. For example, the physical sensation referred to by

    the adjective stiffin (11b) and predicated of Sallys legs may have been felt byher as she woke up the morning after the dancing event took place. Since no

    requirements concerning homomorphism in unsubcategorised object resultativesare in force, any adjective is potentially a suitable one. The situation has been

    schematised in Figure 3.The fact that the two bounded horizontal lines have not been drawn one

    directly under the other indicates the optional character of homomorphism inCase 2 examples.

    Figure 3 Schematic representation of Wechslers (2001)unsubcategorised resultative constructions

    any verb

    any adjective

    homomorphismnot required

  • 7/29/2019 Result at Ive

    9/21

    9

    4. SOME PROBLEMS FOR WECHSLERS (2001) APPROACH

    However appealing Wechslers (2001) model might look, it can be shownthat it runs into several problems when confronted with a large number of data

    (which I have mainly collected from novels and the Internet using the search-engine Google).

    4.1. Impossible classification

    Some examples violate Wechslers classification. Consider the sentences in

    (12):(12) a. They battered him senseless.

    b. Sam cut the bread thin.

    c. He folded this note up very small(J.K. Rowling, 2000,Harry

    Potter and the Goblet of Fire, p. 38)

    They should all belong to Type Ibecause they contain adjectives and verbs

    which are durative (or interpreted as such). (12a) contains the durative verbbatter(see (7b) and (9b)), but a non-gradable scale adjective is used (cf. *verysenseless). The event denoted by cut in (12b) is to be interpreted iteratively

    (hence duratively) but the adjective thin is a non-closed scale one (cf.

    *completely thin). Finally, (12c) cannot be regarded as a Type I resultativeconstruction either. Foldclearly denotes a non-punctual action (hence, Type IIclassification is excluded) but small, although gradable, is not a closed scale

    adjective (cf. *completely small). Further, (12c) cannot be taken as a Type III

    construction since very small can hardly be viewed as the complement of up(and, in any case, up is optional).

    4.2. Context-dependency

    Wechsler (2001) disregards the fact that resultative constructions, more oftenthan not, encode conventionalised (i.e. entrenched) scenarios (see Boas 2000and Verspoor 1997). For example, Wechsler claims that (13a) below

    (13) a. He wiped it {clean/*wet}.

    b. I sprinkled them wet with the garden hose.

    c. *He wiped it to {cleanness/wetness}. (cf. He rubbed his eyes intowakefulness.)

  • 7/29/2019 Result at Ive

    10/21

    10

    is impossible with wet because wet is a gradable minimal-endpoint adjective.

    That is, in the absence of a contextual standard, the standard for wetdefaults to aminimum so that wet is a de facto open-scale adjective (see Wechsler 2001: 5

    for more details). Hence, it is excluded from Type Iconstructions. To be sure,wet appears in resultative constructions with durative verbs, see (13b). If weaccept Wechslers proposal, we conclude that wet in (13b) must be a gradableclosed-scale maximal endpoint adjective. This would mean that the

    interpretation ofwet(as a maximal vs. minimal endpoint adjective) depends on

    the context as suggested by Wechsler. But why should wet not be a maximal

    endpoint adjective in (13a) as well? If Wechslers hypothesis is correct, theremust be some reason for it. I would like to argue that the exclusion of theadjective wet from (13a) does not stem from its contextual interpretation as aminimal endpoint vs. maximal endpoint adjective. Rather, the notion ofexpected

    consequence seems to be the determining factor.

    When we wipe something, we do so because, overwhelmingly, our intentionand/or expected result is that something becomes clean or dry. On the other

    hand, if we use the verb wipe in conjunction with wet, we are resorting to a morecomplex operation, namely one where we negate the expected consequence(s) of

    the action of wiping (viz. cleanness). In some sense, the action of wiping wasnot effective. If we sprinkle something, see (13b), we expect that entity to

    become covered with particles of a certain liquid up to the point in which the

    whole affected entity may be described as wet.Further, one may add the additional problematic fact that both the possible

    adjective clean and the impossible adjective wetin (13a) cannot be replaced by

    the prepositional phrases to cleanness and to wetness (see (13c)), an option

    which should be permitted within Wechslers approach the relevant exampleswould be ofType III.

    Not only do entrenched scenarios motivate why the same adjective (e.g. wet)

    can or cannot appear in a resultative construction, but the very notion of

    gradability as well as the acceptability of allegedly impossible adjectives withinthe same resultative structure are context-dependent. The naturally occurring

    examples in (14a)-(14b) below show that the adjectives triangularand tubular,which are classified as non-gradable by Wechsler and hence regarded as being

    impossible in resultative constructions (see (10c)-(10d)), do pass the very +Adjective test in an appropriate context. (14c) demonstrates that not all speakers

    agree on the impossibility of (10a), *He hammered the metal safe, if morecontext is provided. Hence, we cannot exclude that even the wetvariant of (13a)

    above might be acceptable if a suitable context is supplied.

  • 7/29/2019 Result at Ive

    11/21

    11

    (14) a. A verytriangular roof, the gable allows rain and snow to run offeasily. (www.cmhpf.org/kids/Guideboox/RoofTypes.html)

    b. It [the snake] was verytubular (worm-like).

    (www.nature.net/forums/load/reptile/msg0721102016026.html)

    c. The slide at the park had a section which had come loose. Severalchildren had hurt themselves on the protruding edge. In order toprevent further injuries, John hammered the metal safe. (from

    Verspoor 1997).

    4.3. Additional material

    Adjacent (but not necessarily preceding) material affects grammaticality as isillustrated in (15). Such examples are usually regarded as being colloquial but

    less colloquial than (15a) (unless real is substituted for really in (15d)). Still,

    (15) demonstrates once more that the context of utterance (e.g. formal vs.informal language) should not be ignored.

    (15) a. He painted it nice. (colloquial)

    b. He painted it all nice.

    c. He painted it nice and shiny.

    d. He painted it real(ly) nice.

    4.4. Animacy vs. transitivity

    The notion of homomorphism (between the verbal event and the event hinted

    at by the resultative phrase) does not depend on transitivity, as Wechsler

    contends, but stems from the notion of animacy. In other words, temporaldependency may be optional in subcategorised object cases (see (16b), (17) and

    (18)) and virtually obligatory in unsubcategorised object cases (see (19)).

    (16) a. Sally sprayed her skin wet.

    b. Sally sprayed her skin soft.

    (17) Student stabbed to death. [headline]He was treated by a paramedic and taken by helicopter to hospital, but he

    died soon afterwards. [text] (The Guardian 14.9.1999)

    (18) The critics panned the comedy out of town. (from Levin and Rappaport

    Hovav, 2001)

  • 7/29/2019 Result at Ive

    12/21

    12

    (19) Sally talked her throat dry.

    Whereas the natural interpretation of (16a) is that the event of Sallys

    spraying her skin and that of the skins becoming wet unfold together, this is notnecessarily the case in (16b). The skin may have become soft after the event of

    Sallys spraying it had ended. Such an interpretation relies on our differentconceptualisation of the skin in the two examples under consideration. In (16a),

    the skin is conceptualised as an inanimate surface which was being covered with

    some liquid substance. Hence, homomorphism is expected. In (16b), on the

    other hand, some property intrinsic to the skin actively participates in the eventof its becoming soft. The skin is not conceptualised as an inanimate surface but

    as a (possibly three-dimensional) entity whose properties bring about changes in

    its texture. In this sense, I will say that the skin is animate (although, ofcourse, the skin is not engaged volitionally in the event). Crucially, if some

    property intrinsic to the skin is involved in the process of its change of state (i.e.its change of state does not only depend on external forces), homomorphism

    need not obtain.(17) can be analysed along similar lines. Student is a subcategorised

    argument of the verb stab (and would correspond to the object in the active

    sentence Someone stabbed a student to death). Still, the text makes it clear thatthe student died after the event of stabbing took place, when he was in hospital.

    Processes internal to the students body were of course involved and account forthe lack of homomorphism between the event of the students being stabbed andthe event of the students dying.

    In (18), The critics panned the comedy out of town, the fact that the comedywas moved out of town occurred after the critics had severely criticised it. We

    note that the comedy is of course linked metonymically to the people who areinvolved in it; hence, the lack of homomorphism is once more linked to

    animacy.

    Finally, in (19), Sally talked her throat dry, the interpretation where Sallybecame aware of her throats being dry some time after the event of talkingended is virtually impossible despite the fact that her throat is an

    unsubcategorised object.

    4.5. Construal

    In some cases both a prepositional phrase and an adjective phrase are

    possible although no truth-conditional differences are detectable. Consider (20):

  • 7/29/2019 Result at Ive

    13/21

    13

    (20) a. Afterwards Gonar said we were foolish not to fight Roingin andspear them dead, (Matthew Kneale, 2000,English Passengers,p.53)

    b. The next morning Sutton and two other stock-keepers were found,

    speared to death near their huts. (Matthew Kneale, 2000,EnglishPassengers, p.75)

    Both (20a) and (20b) refer to the actual or potential event of the indigenous

    Tasmanian population attacking Roingin (i.e. the British white men occupyingTasmania) with spears thus causing their death. Suppose that each of the white

    men was speared (not necessarily once) at different points in time and some of

    them did not die immediately. Intuitively, both (20a) and (20b) can code such aninterpretation. According to Wechslers model, however, neither would be

    compatible with it. Since (20a) must be (truth-conditionally) interpretedpunctually (dead is a punctual adjective), one would conclude, adopting

    Wechslers model, that each white man was speared once and died immediatelyboth if there was only one spearing event (i.e. all the white men were attacked at

    the same time) and if there was more than one spearing event (i.e. the white menwere not attacked at the same time). As for (20b), the necessarily multiple

    instances of the spearing event (required by the path associated with theprepositional phrase to death) would refer to the fact that either the white men

    were not attacked simultaneously or they were but each was speared more thanonce and died during the spearing event. In other words, there is no room for the

    interpretation under which not all of the white men were hit repeatedly and died

    during the spearing event. In sum, the difference between (20a) and (20b) doesnot seem to involve truth conditions but, rather, construal, i.e. our ability toconceptualise the same event in different ways by focussing on some of its

    facets at the expense of others (see Langacker 1987). (20a) construes the event

    punctually, by abstracting away from the specific points in time at which eachinstance of the spearing event occurred. On the other hand, (20b) construes theevent as a path with some temporal extension (either because the spearing event

    occurred at different points in time or because the event of dying took some time

    to unfold, i.e. was not instantaneous).

    In conclusion, there is some grain of truth in Wechslers distinction betweenType I and Type III sentences as involving a punctual vs. a durativeinterpretation, respectively. But such a distinction reflects differences inconstrual rather than truth conditions. The use of the adjective correlates with

    viewing the event as punctual (by abstracting away from its actual temporal

  • 7/29/2019 Result at Ive

    14/21

    14

    extension), whereas the use of the prepositional phrase leads us to view theevent as unfolding in real time.

    4.6. Competing forms

    Related to the issue raised in the previous subsection is the observation that

    Case 2 sentences (i.e. unsubcategorised object examples) do not always allowfor the use of an adjective (as we would expect on the basis of Wechslers

    model) when a potential prepositional phrase might also be employed (see (21a-c) and (22a-c)). More generally, we observe that, as happened also with Case 1

    examples, not all conceivable resultative phrases are possible (see (21d-f) and(22c-d)), their acceptability depending for example on the type of verb used (

    see (21e)).

    (21) a. He laughed himself {to death/*dead}.

    b. He laughed himself {to sleep/*sleepy/*asleep}.

    c. He laughed himself {out of a job/*jobless/*unemployed}.

    d. He laughed himself {silly/faint/dizzy/??tired}.

    e. They {laughed/#tittered/#insulted} John out of the room.

    f. #They laughed John {into the room/down the hall}.

    (22) a. He danced himself {to fame/*famous}.

    b. He danced his feet {sore/*to soreness}.

    c. ?He danced himself sore.

    d. *He danced himself crippled.

    (21a) contrasts with (20) above since only the prepositional phrase is allowed.

    However, this might be a matter of construal. Prepositional phrases seem to beused when the consequences of the verbal event (either punctual or not) take

    some time to unfold (see the discussion of (17) above). Now, whereas one candie immediately after being speared or not, the event denoted by (21a) (as well

    as (21b) for that matter) seems to always imply that the change of state tooksome time to unfold. Hence, the prepositional phrase is selected. In sum, I

    propose that

  • 7/29/2019 Result at Ive

    15/21

    15

    (23) Iconicity

    If the span of time required for the change of state to take place isrelatively short, an adjective is used when available (and viceversa).

    Otherwise, a prepositional phrase is used.

    Adjectives focus on the state achieved by some entity (i.e. the affected entity)and, therefore, evoke the temporal path traversed by the affected entity only

    secondarily (i.e. metonymically). On the other hand, dynamic prepositions such

    as to or into profile temporal extension directly. Hence, the punctual nature of

    adjectives iconically accords well with punctual changes and the extendednature of dynamic prepositional phrases iconically accords well with non-

    punctual changes. Significantly, note that my generalisation does not involve the

    nature of the verbal event (punctual vs. non-punctual) and the gradable vs. non-gradable distinction for adjectives, which are essential ingredients in Wechslers

    model. Rather, the choice between adjectives and prepositional phrases (whenboth are potentially available) is a matter of iconic construal.

    We must now account for the fact that two options, i.e. a prepositional phraseand an adjective, are not always available. On the basis of the data in (21) and

    (22), as well as the observation concerning the mono- or bi-syllabic nature of

    resultative adjectives found by Boas (2000) in the BNC (see section 2), Ipropose that

    (24) Phonological length

    If an adjective is used with a prolonged event, then the adjective is

    phonologically shorter than the related prepositions complement.

    (24) captures the contrasts in (21a-c) and (22a-b). All these sentences denote

    change events with some (non-trivial) temporal extension. Hence, giveniconicity, we would expect a prepositional phrase to be used. But the choice

    between an adjective and a related prepositional phrase seems to be regulated byphonological length. In (21a), the adjective dead competes with the related

    prepositional phrase to death. Since dead is not phonologically shorter thandeath (i.e. the prepositions complement), (24) correctly requires the adjective

    not be used. A similar line of reasoning holds for (21b-c) and (22a). (22b) also

    satisfies the phonological length implication because sore is phonologicallyshorter than soreness. Note, however, that, in this case, (24) does not have any

    predictive value; the fact that the adjective is phonologically shorter than therelated prepositions complement does not imply that the adjective is necessarilyused (cf.He rubbed his eyes {awake/to awakeness}). It may be that the choice of

  • 7/29/2019 Result at Ive

    16/21

    16

    the adjective over the prepositional phrase depends, other things being equal, onthe concrete versus abstract nature of the property referred to by the resultativephrase. The impossibility of*She wiped the table to cleanness (at least if uttered

    out of the blue) may depend on the (more) abstract meaning associated with the

    noun cleanness versus the (more) concrete one associated with the adjective.This is also the case with sore vs. soreness, but less so perhaps with awake vs.awakeness.

    Finally, examples like (21d-f) and (22c-d) show that, as was repeatedly

    pointed out above,

    (25) Context-dependency

    The acceptability of a resultative construction depends on how easily thecoded scenario can be activated in the mind of the conceptualiser. In

    general, basic verbs and basic adjectives are preferred since they refer to

    basic (i.e. entrenched), and hence easy-to-activate, scenarios.4

    4.7. Permanent properties

    Contrasts in acceptability do not only obtain between the use of an adjective

    versus a prepositional phrase in the same sentence but also involve the use of thesame adjective in two syntactically similar sentences. Consider (26) and (27):

    (26) a. John painted the fence red.

    b. *John hammered the metal red.

    (27) a. John hammered the metal shiny.

    b. *John chiselled the ice shiny.

    The contrasts in (26) and (27) are not expected given Wechslers model since

    hammer,paint, and chisel all denote prolonged events and the adjectives redandshiny are both gradable and closed-scale (cf. very {red/shiny}, completely

    {red/shiny}). As for (26), the difference seems to involve (relatively) transient

    versus (relatively) permanent properties. In other words:

    (28) Permanent properties

    Resultative adjectives denote (relatively) permanent properties.

    4The diacritics #, ??, * are to be taken as reflecting the progressively diminishing

    acceptability of the examples in question. In other words, even starred sentences could be

    acceptable under appropriate conditions (compare (1a) in 3.1 with (1c) for example).

  • 7/29/2019 Result at Ive

    17/21

    17

    More mysterious is the contrast in (27) since shiny refers to a (relatively)

    permanent property in both sentences. A possible solution will be offered below(see section 5). For the time being, we simply observe that the impossibility of

    (26b) and (27b) is not accounted for by Wechslers formal model.

    4.8. Conceptualiser-oriented adjectives

    Also problematic for Wechslers approach are the sentences in (29), which

    are at best judged colloquial variants of the ones containing the related adverbs:

    (29) a. # Sally painted the room beautiful. (vs. Sally painted the roombeautifully.)

    b. #Sally loaded the cart heavy. (vs. Sally loaded the cart heavily.)

    One could argue that Wechslers model excludes both (29a) and (29b) since

    beautiful and heavy are not closed-scale adjectives (cf. ??completely beautiful,

    ??completely heavy). Still, I have pointed out above that adjective classificationdepends on context (see (14) for example) and hence is not a reliable criterion.

    In fact, I have argued that Wechslers model is untenable as a formal systembecause its very foundation, the prediction of obligatory homomorphism (i.e.

    obligatory temporal dependency) in Case 1 examples, is not correct (cf. (16b),(17) and (18) above). Further, we observe that the adjective beautiful is also

    impossible in a Case 2 example like (30a), about which Wechslers model hasnothing to say.

    (30) a. *He danced his legs beautiful. (intended interpretation: His legs

    became beautiful as a result of his dancing)

    b. He danced his legs sore.

    I contend that the deviance of (29) and (30a) can be captured by the

    following generalisation:

    (31) The experiential generalisation

    Resultative adjectives (in non-colloquial usage) refer to non-aesthetic(e.g. She painted the room {red/*beautiful}), necessarily perceivable (e.g.

    She hammered the metal flatvs. *She loaded the cart heavy) properties.

    Whereas the adjective red in She painted the room red describes a non-

    aesthetic (i.e. objective) property of the room, beautiful in (29a) and (30a) does,of course. Note also that beauty is attributed to the affected object by the

    conceptualiser, but is not out there in the real world (whereas we assume

  • 7/29/2019 Result at Ive

    18/21

    18

    that colours are). As in the well-known idiom Beauty is in the eye of the

    beholder, beauty does not exist independently of us (or at least we think so).Heaviness in (29b) is not necessarily directly perceivable (but see (32) below)

    and might only be verifiable indirectly by using, for example, a pair of scales

    with which to measure the weight of the cart with the crates on it. Interestingly,it is possible to find instances where heavy does occur in contexts similar to(29b). Let us analyse the following example:

    (32) I once loaded it [i.e. the truck] heavy on the back end and experiencedsome minor rubbing (only on big bumps) but have had no other problems

    in over 40K miles. (http://www.ford-trucks.com/dcforum/earlybronco/20.html)

    Although the use of heavy might be colloquial and/or influenced by theoccurrence of additional material (see section 4.3), one cannot help noticing that

    the consequences of the heavy load on the truck are visible (i.e. directlyperceivable) in (32). The truck rubbed on (big) bumps on the road, that is we

    understand that the back of the truck was in a lower position than usual.

    Finally, it must be stressed that (30b), which differs minimally from (30a), is

    compatible with the experiential generalisation. The sensation of soreness,although not an objective property in the same sense as flatin Sally hammered

    the metal flat, is nevertheless experienced by the subject referent and not by the

    conceptualiser (who is external to the event). In this sense, I say that sore in(30b) refers to a perceivable property (although, unlike colours, not a visualone).

    5. GESTALT VERSUS PART-WHOLE PROPERTIES

    So far I have shown that Wechslers model does not make correct predictions

    and cannot be maintained because various examples belies its very foundation,the homomorphism requirement for subcategorised object cases. I have argued

    that a series of factors contributes to the acceptability of adjectival resultativeconstructions, such as iconicity, phonological length, context, the distinction

    between permanent vs. temporary properties, and perception. Nevertheless, Ihave not accounted for all cases yet. The contrast between (33a) and (33b) is not

    captured by any of the criteria I have proposed, as the reader can easily verifyfor herself or himself.5

    5For example, it is difficult to say whether (1a) codes a more entrenched scenario than (1b)

    (especially as far as the variant with the adjective long is concerned).

  • 7/29/2019 Result at Ive

    19/21

    19

    (33) a. He hammered the metal flat.

    b. *He hammered the metal {long/tubular}.

    I propose that the crucial notion that allows us to explain the contrast in (33)

    is the distinction between part-whole properties and gestalt properties. In moredetail:

    (34) The part-whole affectedness generalisation

    If an adjective describes a property P of an affected object Y, then P

    describes any part of Y (if possible).

    Whereas the adjective flat can be predicated of every part of the affected

    entity metal (e.g. the metal cannot have any bumps on it), long and tubularonly

    refer to the metal as a whole (i.e. to its overall shape). We can select parts of it

    which are neither long nor tubular (e.g. provided an appropriate reference lengthsuch as 30cm, a piece of metal 1m long could be described as long but onehundredth of it, that is 1cm, cannot). Adjectives like flat will be called part-

    whole adjectives, whereas adjectives like long and tubularwill be referred to asgestalt adjectives.

    Notice that a sentence like (35)

    (35) He knocked the chair crooked.

    is not a counterexample to the part-whole affectedness generalisation. The

    property of being crookedapparently only refers to a part of the chair (e.g. oneof its legs). However, the fact that a part of the chair has a leg which is shorter

    implies that every part of the chair is no longer straight with respect to itsoriginal position. In other words, the change in one of its parts implies that all

    other parts of the chair are altered as well with respect to the spatial axes. Thisalso might mean that crookedin reality refers to a spatial configuration rather

    than a property. If this is so, (35) is not dealt with by the part-whole affectedness

    generalisation. In any case, it is clear that no spatial rearrangement for all partsof the metal is implied in cases like (33b), *He hammered the metal{long/tubular}.

    The part-whole affectedness generalisation also covers sentences like Chriscut the bread thin, where the affected object is the breadand its parts are theslices (each of which is thin).

    The qualification in brackets in (34), if possible, is intended to account for atleast two cases where part-whole adjectives are not used. First, the affected

  • 7/29/2019 Result at Ive

    20/21

    20

    entity may be conceptualised as a point. For example, in Sally battered Chris

    senseless, senseless can only describe the affected entity as a whole; it does notmake sense to predicate it of a part of the affected entity. That is, Chris can be

    regarded as a point rather than an entity with some spatial extension in the

    present context. Similarly, in Milton read himself blind, blind can only bepredicated of Miltons eyes, not of any other part of his body.

    Second, the verb, as is employed in the construction, may only refer to a

    property of the whole affected entity (i.e. the affected entity does have spatial

    extension but such extension cannot be targeted for predication). Consider (36):

    (36) I saw him coming back, carrying two sacks that were heavy so theypulled his arms long. (Matthew Kneale, 2000,English Passengers, p.258)

    Long is impossible in resultative constructions like (33a) above, He

    hammered the metal flat, because it describes a property of the whole object butcannot be used to describe any arbitrarily chosen part of the metal. On the other

    hand, long does occur in a resultative construction in (36). Of course, it could be

    argued that long in (36) is interpreted metaphorically, standing for a spatialconfiguration (rather than a property) such as down6. This would imply that (36)lies outside the jurisdiction of the part-whole affectedness generalisation.

    Alternatively, we may note that the verb pull in (36) necessarily refers to an

    action carried upon an entity (i.e. his legs) as a whole: whereas one can hammer

    parts of a metal, one cannot pull parts of an arm (under the intendedinterpretation of (36)).

    Finally, it is worth pointing out that cases like They piled the crates high do

    not violate the proposed generalisation because there is strictly speaking noaffected object. Rather, we have an effected object since a pile (of crates) is

    created. The use of the adjective high is however compatible with theexperiential generalisation because high denotes a necessarily perceivable

    property (i.e. the visual property height).

    6. CONCLUSION

    In this paper I have argued that the most recent approach to adjective

    selection in resultative constructions, i.e. Wechslers (2001) model, runs into a

    6This is also the case with sentences such as He opened the door wide, He closed the door

    shut,He broke the walnut apart, where the adjectives do not refer to properties of the affected

    objects door and walnut but, rather, describe (created) spatial configurations: the length

    between the door and its frame (as seen from the front) was great or null, the walnut was

    divided into two parts.

  • 7/29/2019 Result at Ive

    21/21

    21

    series of fatal problems. In its place, I have proposed various generalisations,among which are the experiential generalisation and the part-whole affectednessgeneralisation. They are compatible with the large number of data I have

    considered and cover all adjectival cases found by Boas (2000) in the BNC. I

    have also stressed that resultative constructions often code entrenched (i.e. easy-to-activate) scenarios. Such an observation correlates well with the limitednumber of adjectives found in the BNC. They are usually mono- or bi-syllabic

    and also refer to basic (i.e. entrenched) aspects of life (e.g. death vs. life,

    positions, quantity, basic physical/psychological states).

    REFERENCES

    Boas, Hans. 2000. Resultative Constructions in English and German. Ph.D. Thesis, University

    of North Carolina.

    Broccias, Cristiano. 2003. The English Change Network. Forcing Changes into Schemas.

    Berlin: Mouton de Gruyter.

    Haegeman, Liliane. 1994. Introduction to Government and Binding Theory. Oxford:

    Blackwell.

    Kneale, Matthew. 2000. English Passengers. London: Hamish Hamilton.

    Langacker, Ronald. 1987. Foundations of Cognitive Grammar. Vol. 1: Theoretical

    Prerequisites. Stanford: Stanford University Press.

    Levin, Beth. 1993. English Verb Classes and Alternations. Chicago: University of Chicago

    Press.Rappaport Hovav, Malka and Beth Levin. 2001. An event structure account of English

    resultatives. Language 77: 766-797.

    Rowling, J. K. 2000. Harry Potter and the Goblet of Fire. London: Bloomsbury.

    Verspoor, Cornelia. 1997. Contextually-Dependent Lexical Semantics. Ph.D. Thesis,

    University of Edinburgh.

    (Available at www.cogsci.ed.ac.uk/~kversp/ftp_html/thesis.html)

    Wechsler, Stephen. 2001. An analysis of English resultatives under the event-argument

    homomorphism model of telicity. Proceedings of the 3rd Workshop on Text

    Structure. University of Texas at Austin.

    (Available at http://uts.cc.utexas.edu/~wechsler/papers.html)