Modern Grammars of Case anderson phần 2 pdf

But his suggestion concerning the status of ‘grammatical relations’ reverses the traditional conception of the relationship between the relations and the structural properties that signal their presence. The whole burden of the classical tradition concerning case, adposition, and word order is that the relations cannot be identiWed with or by their manifestations. Despite the relations allowed for by Chomsky apparently being superXuous in his framework, problems with such deWnitions have been much discussed. Issues include: the invocation, as already alluded to, of two sets of relations, ‘deep’ and ‘surface’, which involve diVerent kinds of deWnition (Chomsky 1965: ch. 2,n.32); the questions raised by ‘double-object’ constructions; the intractability of ‘non-conWgurational’ languages and other indications that subjecthood may have a more independent role; and the possible non- universality of these relations. For a detailed contemporary critique see for example Dik (1968:§8.2). We return to some of these at various points later. But perhaps most relevant to our present concerns is the observation that the grammar of Chomsky (1965) is inconsistent in its treatment of relations. Included in the ‘illustrative fragment of the base component’ given in Chomsky (1965: ch. 3,§3) are the rules in (18): (18) a. Predicate-phrase ! Aux ^ VP (Place) (Time) b. VP ! V (NP) (Prep-phrase) (Prep-phrase) (Manner) c. Prep-phrase ! Direction, Duration, Place, Frequency, etc. The rulesprovided do not oVer anexpansion ofthe ‘Place’ and ‘Time’ categories of (18a); but presumably one possibility is (19), in the light of the likes of (20): (19) Place, Time ! Prep-phrase (20) We shall sell wine in Adelaide in the future By combined application of (18)and(19 ), Place, at least, both immediately dominates and is immediately dominated by Pr ep-phrase, as can be seen in (21): (21) a. Place (by (18a) þ (19)) j Prep-phrase b. Prep-phrase (by (18b) þ ( 18 c)) j Place This is one kind of inconsistency, one that is damaging enough. But it is clear that yet another, more fundamental inconsistency underlies this one. 32 Modern Grammars of Case For the alternative expansions in (18c) all involve relations, in this case semantic relations (as observed by Fillmore (1965)). Within the (autono- mous) syntax, categorial recognition is given, paradoxically, to semantic but not syntactic relations—though it is not acknowledged in Chomsky (1965) that this is what is involved. And the primary motivations for including these categories are apparently semantic; they are given no syntactic motivation, at any rate. One problem here is that the semantic distinctions among Place and the rest are signalled by prepositions and the lexical items in the noun phrases that complement them, even though, according to (18c), the preposition is in this case outside the Place phrase itself. The semantic properties of Place are carried by elements outside the Place phrase. This suggests that this is not the way to characterize semantic relations grammatically—by the back door, as it were. And this failure is a consequence of an attempted autonomist attitude: syntax does not invoke semantics; the syntax of syntactic elements is inde- pendent of their semantics. Moreover, in this particular instance, the semantic properties associated with Place, Direction, etc. do have syntactic consequences. Thus in German adverbs of Time and Manner normally precede those of Place, as in (22): (22) a. Er geht jetzt nach Hause he goes now to home (‘He is going home now’) b. Er fa ¨ hrt mit dem Zug nach Hause he travels with the train to home (‘He goes home by train’) One can exclude such phenomena by Wat from ‘syntax’; but this is a rather transparent device to avoid disconWrmation of the autonomy hypothesis at the expense of the traditional understanding of the domain of syntax. In these instances, of course, one might appeal to the complement/ adjunct distinction (adjuncts before complement in (22)), but that distinction too is semantically based (§2.2.1). Moreover, there are other instances of ordering involving adjuncts of diVerent semantic classes, as established by Frey and Pittner (1998; 1999). Thus, the comitative in (23) precedes the instrumental: (23) a. Er hat (zusammen) mit einem Freund mit einem he has (together) with a friend with a Kleintransporter den Schrank herbeigeschaVt minivan the wardrobe hither.brought The Classical Tradition and its Critics 33 (‘He (has) brought the wardrobe here with (the help of) a friend in a minivan’) b. (?) Er hat mit einem Kleintransporter (zusammen) mit einem Freund den Schrank herbeigeschaVt As Frey and Pittner (1998) put it, (23), their (56a, b) shows ‘dass man Instrumentaladverbiale lieber beim Verb plaziert als Komitative [ ] Diese semantische Pra ¨ ferenz hat aber keinen syntaktischen Niederschlag im Sinne von syntaktisch Wxierten Grundpositionen, die die Adverbiale relativ zuei- nander einnehmen wu ¨ rden.’ There are reasons for thinking that Place and Time, for instance, are composites (phrases) that have in common their relational structure: rela- tionally, they are both instances of location in diVerent dimensions, as is discussed later. It’s the substance of the dimension rather than the relation in which they diVer; and for (22) it is the former which is relevant to the syntax. But such relations themselves also have a syntax, as well as these composites. Let me give a preliminary illustration—though what follows will illustrate this in some depth. It is not implausible to distinguish between a semantic relation Source and a semantic relation Path, presumably subtypes of Chomsky’s Directional. These are respectively illustrated in (24a–c) and (d)—where we’re not con- cerned with whether the post-verbal phrases are complements or adjuncts (where the notation of (24c) indicates that the conjunction cannot be omitted): (24)a. * Fred came from Birmingham out of the Midlands/North b. * Fred came from Birmingham and out of the Midlands/North c. People came from Birmingham * (and) from Leicester d. Fred came through the valley (via Stirling)(and across the plain) Comparison of (24a, b) with (d) shows that only the Path can be duplicated, with or without coordination; the Source cannot be duplicated if a single journey is involved, whether or not the second Source is coordinated with the Wrst (24b) or not (24a). (24a, b) are either tautologous (Birmingham, England, is in the Midlands) or contradictory (Birmingham is not in the North). Of course, the Source can be coordinated, without diYculty if this is compatible with the subject and aspect, as in (24c); but an un- coordinated version is not available. All of this ties syntactic possibilities to semantic relations—as assumed, amongst other things, in a grammar of case. 34 Modern Grammars of Case 2.3 Conclusion From what I have described here concerning early transformational grammar, one gets the impression of a grammar of case struggling to escape from the bonds of Chomsky (1965), as implied already by Matthews (1967: 132–5). Chomsky’s book illustrated that it was a time to return to grammars of case, to explore the aspirations of the philosophical grammarians, so mis- leadingly extolled by Chomsky. Thus, more generally, what I’ve said so far is intended to illustrate something of the extent to which what happened at this point was a return. And this is already apparent in terms of the background invoked in Anderson (1971b), as well as in the brief survey of recent ideas on case oVered by Fillmore (1968a). American structuralism and its transform- ational oVshoot, on the other hand, represented a signiWcant (and, I shall argue, unnecessary) break with ‘traditional grammar’. The Classical Tradition and its Critics 35 3 Early Case Grammar A return to a concern with grammars of case was attempted in a variety of work in the late 1960s and 1970s, some of which came to be referred to as ‘case grammar’. What further distinguishes what it seems to be appropriate to label ‘case grammar’ from the main tradition of grammars of case, however, is the adoption of a more restrictive and explicit view of the relationship between semantic and grammatical relations than is embodied in a grammar of case of level 2. A level 2 grammar, indeed, doesn’t specify any relationship between the diVerent kinds of relation. Such a development is Wrst evident in the work of Fillmore (1965; 1966; 1968a; 1971), but it is also manifested, independently to begin with, in a range of other contemporary work, including Anderson (1968a; 1969; 1971b), Brekle (1970), and Chafe (1970), as well as to some extent in variants of ‘functional grammar’. I shall concentrate to begin with on Fillmore’s work, which for many people, particularly in the USA, came to be identiWed with ‘case grammar’, and which will help us, I think, to establish what is distinctive about what we might want to call ‘case grammar’. What comes to be envisaged at this point is what I shall refer to as a grammar of case of level 3: Grammar of case level 3 a) A grammar of case gives an account of the syntax of the relations that are typically expressed by case inXections or adpositions or position. b) Among these relations semantic relations have primacy. This formulation adds to that for grammar of case level 2 0 (¼ (a) here) the stipulation (b) that semantic relations have primacy over grammatical. ‘Pri- macy’ can be articulated in various ways, of course, depending on other properties of the grammar; the determination of how it is to be articulated constitutes one unWnished task of early ‘case grammar’, or at least one that came to be seen as not satisfactorily concluded. However, assertion of the ‘primacy’ in principle of the semantic over the grammatical relations distin- guishes ‘case grammar’ from most other grammars of case. On the other hand, it seems to me that ‘case grammar’ is in this respect more clearly than ‘core’ transformational-generativ e grammar the ‘spiritual heir’ of the tradition of philosophical grammars to which the work of the Port-Royal group belongs. I have contended that the work of that group is not a prelude to the central transformational tradition; such a view is the result of wish-fulWlment based on perceived shared ideas in the philosophy of mind rather than reXecting comparison of grammatical analyses in the two traditions. As I have suggested, the work of the Port-Royal group incorporates a grammar of case of at least level 2 0 . And I now suggest that they seem to have in mind a grammar of level 3, as just formulated, though this is not precisely articulated. And I think this emerges clearlyeven fromjust acomparison of the quotation concerning Port-Royal from Chomsky given in §2.1.4 with what follows. But for a more detailed basis for comparison, see for example Lancelot and Arnauld (1660:chs.7, 11, 12). 3.1 The Fillmorean initiative Against the background of transformational-generative grammar, Fillmore (1965; 1966; 1968a) suggested a transformational relationship between (struc- tures containing) semantic relations, on the one hand, and (structures asso- ciated with) grammatical relations, on the other. Gone are the ‘deep’ grammatical relations of Chomsky (1965); instead, the identiWcation of ‘sur- face’ grammatical relations is based on structures derived from ‘underlying’ structures which crucially contain nodes corresponding to diVerent semantic relations; there are only ‘surface’ grammatical relations. Even at this point we can be a little more speciWc about the form of the grammar proposed by Fillmore, which can be spelled out as follows: Fundamental concepts of case grammar a) The constructional relevance of semantic relations There is a level of syntactic structure that is constructed on the basis of (among other things) the semantic relations contained in the lexical entries of predicators. b) The irrelevance of ‘deep structure’ This level replaces (and displaces) ‘deep structure’ as the interface with the lexicon and as basic to syntactic structure. This attempts to highlight two of the concepts that most clearly serve to distinguish early ‘case grammar’ from the main transformational tradition. Let us look in more detail, and more formally, at what these ‘fundamental concepts’ involve. Early Case Grammar 37 3.1.1 ‘Cases’ and grammar Thus, Fillmore suggests ‘underlying’ structures like the respective representa- tions given in (2) for the sentences in (1): (1) a. The door opened b. The girl opened the door (cf. Fillmore 1968a: §3.1). The S(entence) consists of a M(odality) constituent and a P(roposition); the former is realized by various sentential modalities, including tense, and the latter contains the verb and its arguments. O and A are semantic relations, what Fillmore calls ‘case relations’, or simply ‘cases’. I shall return in due course to the identiWcation of the semantic relations O(bjective) and A(gentive). K is a ‘kasus’, a non-phrasal category that in English is usually realized by a preposition; a preposition, such as by in (2b), is thus (whatever else) a kind of ‘case form’. But we can already see (2) S M P : : V O : : : : : : : : K NP : : d N : : :: : : Past open the door b. S M P : : V O A :: : : K NP K NP : : : : : D N : DN :: : : : : : : : : : : : : Past open the door (b y ) g irlthe a. 38 Modern Grammars of Case from these representations that apparently not all instances of ‘cases’ are marked by a preposition. I take this up again in a moment. Part of the restrictions on what arguments a verb can take is speciWed in terms of the set of semantic relations it can co-occur with in a proposition, its ‘case frame’. In this instance we say that the frame for open is at least as in (3), where the brackets enclose an optional element, one not associated with all instances of open: (3) open O (A) (I’m not concerned at this point with whether (3) is exhaustive or not.) The content of the ‘case’ node percolates down into its kasus and noun phrase constituents; and this may be reXected in, for example, choice of preposition, so that by is, among other things, a marker of A. The elements in both (2) and (3) are not ordered in sequence; the trees in (2) are ‘wild trees’ (Staal 1967). This comes close to being another ‘funda- mental concept’, but one that was not fully exploited until somewhat later in the development of ‘case grammar’, as well as not as distinctive of ‘case grammar’ as concepts (a) and (b). We shall return to begin to look at the signiWcance of this further ‘concept’ in §3.1.2. On Fillmore’s account, only after the conWgurations associated with gram- matical relations are created is order determined, as in (4): (4) S NP M P : | d N : V : : : : : : : : the door Past open a. b. S P V : : : : open M : : : : : Past PN d N : : : : the girl PN d N : : : : the door Early Case Grammar 39 The subject in (4a, b) is selected in terms of a hierarchy of ‘cases’ in which A outranks O as subject choice. The ‘passive’ in (4c) involves a ‘marked’ selection of subject triggered by the presence of [þPass] on the verb, and signalled as ‘marked’ by the presence of the be and the participle form. Formation of subject and object involve pruning of the case (and kasus) nodes as well as positioning of the remnants. There are (by now familiar) problems with the kind of analysis of passives proposed here. In the Wrst place, it involves attributing to transformations the power, allowed to early transformational grammars, of introducing lexical material, here be. This is undesirable on theoretical and (in this case) empir- ical grounds. It greatly enhances the already undesirable power of transform- ations; and it undermines any claim that it is only at ‘deep structure’, or rather its replacement in ‘case grammar’, that there is access to the repository of lexical material, the lexicon. Later developments of passive in ‘case grammar’ (as described in §§9.2, 12.2.2)oVer interpretations compatible with more restrictive syntactic assumptions, and more in accord with other observations concerning the syntax of passives, as anticipated in what immediately follows. Secondly, it is the presence of be that determines the morphological form of the other verbal; normally such rection or government of morphological form goes from an element to its complement, as with the preposition and pro- noun in (5), where the complement is underlined: c. S P V [+Pass] : : : : : open M PN d N : : : : : : : : the Past be door A K : : : : by NP dN : : : : the girl ð5Þ She went towards him 40 Modern Grammars of Case It’s not clear in Fillmore’s account that the non-Wnite can be described as the complement of passive be, particularly when the latter is part of a complex M, as in (4c). Finally here, the representation in (4c) does not express the adjunct status of the by-phrase, which as such is optional; the sentence is complete without it, it is intransitive. In many languages such a phrase is preferably or indeed obligatorily absent. However, these are problems shared by the standard transformational analyses of the time. They are problems shared by any analysis that doesn’t recognize that the passive contains two verbals, with the non-Wnite verb form subordinate to the ‘auxiliary’, and by an analysis that fails to recognize that the valency of the passive participle is diVerent from that of the other verb forms. Resolution of the problems doesn’t introduce a problem for the subject selection hierarchy of ‘cases’; indeed, it is supportive thereof. For instance, if the non-Wnite in passives is ‘intransitive’, i.e. it does not take an A comple- ment, there is nothing to outrank the O in this case. And in this way the claims embodied in the ‘case grammar’ concepts (a) and (b) remain intact. 3.1.2 Linearity I should say something at this point about the absence of linear precedence relations from (2) and (3). Anderson (1971b) also assumes no initial ordering; underlying structure involves ‘wild trees’, as envisaged by Curry (1961) and S ˇ aumjan and Soboleva (1963). And ordering is derived on the basis of other information, and imposed more superWcially. This view is defended in Anderson (1977:§1.11), particularly against objections raised by Chomsky (1965: 124–6 ) and Bach (1975). Bach argues for a universal base which attri- butes the same order to the elements of all languages, but fails to make a convincing case on either empirical grounds or grounds of restrictiveness: he claims that this hypothesis ‘rules out more possible states of aVairs within its domain of application’ (1975:§1), which is simply not the case. Such a (‘universal underlying word order’) hypothesis continues to be raised in various forms, but not in any way that disturbs the conclusion of Koutsoudas and Sanders (1974: 20): The most restrictive and empirically most well-supported hypothesis about constitu- ent ordering is in fact that which asserts that all underlying representations are wholly free of ordering speciWcations, that such speciWcations are assigned by rules to the superWcial groupings of superWcial constituents, and that all ordering relations are derivationally invariant. (Cf. too Sanders 1970; 1972.) Early Case Grammar 41 [...]... proposals in particular of two fundamental concepts of case grammar’, formulated as: 54 Modern Grammars of Case Fundamental concepts of case grammar a) The constructional relevance of semantic relations There is a level of syntactic structure that is constructed on the basis of (among other things) the semantic relations contained in the lexical entries of predicators b) The irrelevance of ‘deep structure’... problem 3 .2. 1 Dependency Robinson (1970b) and Anderson (1971a) argue for characterizations of semantic relations in terms of dependency relationships (in the sense of the ` tradition(s) of Tesniere (1959) and Hays (1964)) In terms of the framework of Anderson (1971b) (though retaining Fillmorean labels—and see for example Anderson (1977: 2. 2) for a much fuller discussion), we should reinterpret (2b) as... in the subsequent discussion, as we see ideas of representation evolve in 46 Modern Grammars of Case various ways in relation to other developments, particularly in response to (11iii) and (iv) Chapter 4 explores the consequences of (11ii) 3 .2 The representation of case relations and forms Fillmore himself was uncertain about the proper representation of case (1968a: 87); and many workers using his... 2. 1 .2; 1998a: 2. 2.1; Anderson 1991: §§4–5; 1997: §§1.3, 3.1): Inalterability condition The relations of dependency and sequence assigned to an element are inalterable Much of the elimination of movement suggested in (for instance) Anderson (1977) involves (re)interpretation of alleged movements as reattachment What subsequent work adds, in the light of inalterability, is the reinterpretation of ‘reattachment’... this is the head of the construction in terms of (13/14)—where percolation is unnecessary, because case and kasus are identiWed 3 .2. 2 The categorial identity of case and preposition: a functional category Another motivation for the adoption of dependency representations came from the proposals made at about the same time in Anderson (1971c) These provide dependency representations of case with the... implementation of this lexical mechanism again awaits later developments in the theory of categories, discussed here from Chapter 8 onwards 3.3 Conclusion Much of linguistic structure (in particular) remained underspeciWed in these early case grammar’ proposals What united them was an acceptance of the goal of a grammar of case of level 3, as formulated initially in this chapter, i.e.: Grammar of case level... obviously the introduction into the syntax of semantic relations, in the shape of Fillmore’s Early Case Grammar 45 ‘cases’ Fillmore (1968a: 24 –5) oVers the deWnitions in (10) for the three ‘cases’ we have been looking at so far: (10) Agentive (A), the case of the typically animate perceived instigator of the action identiWed by the verb Dative (D), the case of the animate being aVected by the state... by most contributors to the case grammar’ programme 4 Case Grammar and the Demise of Deep Structure The suggestions made in Chapter 3 go some way towards the development of a framework that conforms to the formulation of a grammar of case of level 3 But many questions remain; some of the more important are embodied in (iii) and (iv) of (3.11), just repeated at the end of Chapter 3 Let us look Wrst,... do with Aktionsart, and do not correlate neatly with the phenomena in (20 , 21 ); and the latter are ¨ compatible with a case grammar’ interpretation, as we have seen Bohm (1993: §4 .2. 2) illustrates the importance of Aktionsart for ‘perfect-auxiliary’ selection with the pair of ‘perfects’ in (22 ) containing the same German verb: (22 ) a b Molly hat auf Molly has on der the:DAT ¨ Buhne stage:DAT getanzt... stage:ACC danced (22 a) represents a ‘process’, it is ‘atelic’; (22 b), on the other hand, is an ‘accomplishment’, and ‘telic’ (see also Tenny 1994) Sorace (20 00) and Keller and Sorace (20 03) illustrate the complex and gradient nature of the factors inXuencing choice of auxiliary in Italian and German respectively They establish a hierarchy of ‘intransitive’ verbs, from Case Grammar and the Demise of Deep Structure . shall refer to as a grammar of case of level 3: Grammar of case level 3 a) A grammar of case gives an account of the syntax of the relations that are typically expressed by case inXections or adpositions. (Bo ¨ hm 19 82: §§1.1 .2, 2. 1 .2; 1998a: 2. 2.1; Anderson 1991:§§4–5; 1997:§§1.3, 3.1): Inalterability condition The relations of dependency and sequence assigned to an element are inalterable. Much of the. replacement in case grammar’, that there is access to the repository of lexical material, the lexicon. Later developments of passive in case grammar’ (as described in §§9 .2, 12. 2 .2) oVer interpretations