Agnates, verb classes and the meaning of construals ...

9 downloads 0 Views 3MB Size Report
Eynde's distinction between 'dispositive', or systematic, clausal variants and ..... constellation, while (20a) should be analyzed as Actor-Goal-Place:motion.
Leuvense Bijdragen, 87(3-4), 1998, 281-313.

Agnates, verb classes and the meaning of construals. The case of ditransitivity in English. Kristin DAVIDSE K.u. Leuven

O.

Introduction

The 'Syntactic Variation' project has continually confronted all its researchers with theoretical and methodological questions about the status of paradigmatic variants of constructions. Some participants in the project were situated in the functional research tradition which handled paradigmatic variants in terms of Gleason's concept of 'agnates'. Others had a background marked by Van den Eynde's valency methodology in which 'reformulations' and 'linked constructions' constitute the main methodological tool. Thence grew the twofold aim of this article. Firstly, it reflects on the theoretical importance and heuristic value accorded by these research traditions to paradigmatic variants, as a means of identifying construction types and delineating verb classes. Secondly, it illustrates some of these methodological issues with ditransitive construals and ditransitive verbs. This is, of course, an area of English lexicogrammar displaying the variation between bare nominal and prepositional phrases that the 'Syntactic Variation' project is concerned with. The first half of this article (section 1) will be devoted to theoretical and methodological reflection. In section 1.1, the view will be advocated that relations between constructions are essential to a sound conception of the language system and linguistic categories. The main theoretical pathfmders in this area were Gleason, with his concepts of enation and agnation, and Whorf, with his distinction between phenotypical and cryptotypical categories. In section 1.2, I will look at the role played by agnates in the identification and interpretation of construction types. Firstly, I will reflect on the meta-functional diversification which is imposed on relations between constructions in a Hallidayan perspective. I will then link this to Vanden Eynde's distinction between 'dispositive', or systematic, clausal variants and selective, verb-specific, re-constructions. In section 1.3, I will look at how EGGERMONT (1994) interpreted Van den Eynde's principles to arrive at a language-based, non-intuitionist, classification of French verbs. - 281 -

K.DAVIDSE

In the second half of this article (section 2), I will apply paradigmatic reasoning to the area of ditransitivity and show how, both in interpreting the semantics of constructions (section 2.1) and (sub)c1assifying verbs (section 2.2), it leads to different results from those found in syntagm-based mainstream approaches. In this way, I would also like to invite more explicit discussion of the methodological and theoretical issues which are at the heart of this article.

1.

Relations between constructions : reflecting about the language system and Hoguistic categorization

1.1.

Gleason on enation and agnation

Looking at the development of modern linguistics in the fITst half of this century, Firth expressed the general criticism that it had neglected the paradigmatic at the expense of the syntagmatic, in spite of Saussure's balanced treatment of both axes. By and large, mainstream linguistics in the second half of this century has still tended towards a syntagmatic bias. However, one linguist who should certainly be singled out for having reacted against this bias is Henry Allan Gleason. At the time in linguistic history when transformational grammar was emancipating itself from American structuralism, GLEASON (1966) assumed an original theoretical position. He stressed (1966: 195) that the language system is not simply an inventory of syntagmatic structures - as some forms of structuralism had held - but that it also crucially involves the paradigmatic relations between structures. These, GLEASON (1966: 195) said, constitute "an additional set of relationships that grammar must describe [...J at least as large and diverse" as the syntagmatic patterns. At the same time, he was at pains to stress that the relations existing between structures as an inherent part of the language system could not be captured by transformations. Transformations are operations realizing 'moves' through the system, but do not describe its inherent properties. GLEASON (1966:196) put this in the following simile : "A trip is not part of a highway system, though only the highway system makes it possible." To get a grasp of the relations between structures, GLEASON (1966: 199) proposed two new technical terms, enation and agnation. Enation is the relation of structural identity, obtaining between examples which have an identical structure relating members from identical grammatical classes. For instance,

- 282-

Leuvense Bijdragen 87(1998)

(1) The cat broke the glass. (2) Lizzie opened the door. (3) Malicious rivals spread rumours.

Agnation is the relation of systematic and. regular grammatical variatio n existing between examples whose main lexical elements are identica l. For instance, (4) a (4) b

The cat broke the glass. The glass was broken by the cat.

Gleason stressed that enation and agnation are two mutually defining notions. On the one hand, the systematicity of an agnation pattern should be confirmed by enate examples displaying the same agnation relation. For instance , (5) a (5) b

The dog bit the man. The man was bitten by the dog.

(6) a (6) b

The vicar hit the burglar. The burglar was hit by the vicar.

On the other, for examples to be enate, they should have identica l sets of agnates. Gleason also pointed out that if examples shared some agnates but not all, they could be 'partiaU y enate'. It seems consistent with this way of thinking to posit that examples which share hardly any or no agnates are 'non-en ate', even though Gleason himself never used this concep t By defming structural identity in terms of shared agnation, Gleason adduced a vital method to detect structural ambigu ity. For instance, despite the superficial similarity in structure between (7) The man saw a stranger

and (8) The man seemed a stranger,

these two sentences are not enate, as shown by their different agnate relation s

(GLEASON 1966:203).

(7) a (7) b

A stranger was seen by the man. *The man saw to be a stranger.

(8) a (8) b

*A stranger was seemed by the man. The man seemed to be a stranger.

The methodological importance of this point is paramount - and remains controversial to this day. From the perspective of agnation heuristics, main-

- 283-

lj

I

\

i

K.DAVIDSE

I

~.

stream descriptions can be said to still group together as structurally identical some construction types whose different agnation paradigms in fact show them to be distinct structural configurations. While agnation reasoning has been used on an ad hoc basis by most linguistic schools, its systematic application has remained decidedly non-mainstream. The most important point of controversy revolves around what counts as formal evidence. In a Gleasonian perspective, formal evidence is not restricted to 'observable characteristics' of the syntagm in question, but it also includes the syntagm's systematic relation to its agnates. Or to put it in more mainstream terms, it includes the syntagm's behaviour with regard to syntactic tests. Refusal to recognize and systematically apply such paradigmatic reasoning has been motivated, amongst others, by claims that it would render falsification of descriptive hypotheses vacuous, that 'anything would go'. While one may want to reject or restrict paradigmatic arguments for other reasons, I do not think this one holds. The empirical facts of language in use present us with clear 'can' s and 'cannot's' (GLEASON 1966), i.e. possible and impossible agnates. These lend themselves as well to verification and systematization as the overt characteristics of the isolated syntagm. Finally, it should be noted that Gleason viewed agnation not only as a means of formally identifying distinct constructions, but that he had also noted the relevance of agnation to the semantics of a construction. As he observed (1966:213), a more explicit agnate construction is often felt to reveal the meaning of the initial construction. For instance. an automatic lathe operator 'means' an operator of automatic lathes, not *an operator of lathes that is automatic. Still, Gleason's work does not contain very explicit reflection on agnates as a semantic heuristic nor on their relevance to the identification of verb classes. Arguably, some of Whorf's reflections about cryptotypical categories are more suggestive in this regard. 1.1.2. Whorf on phenotypical and cryptotypical categories WHORF (1956:87-101) made a fundamental distinction between phenotypical and cryptotypical categories. Phenotypes are overt grammatical categories : they are realized by one systematic formal mark, such as a specific morphological inflexion or a certain syntactic pattern. Case as a morphological category is a classic phenotype. Such overt categories lend themselves well to a 'realizational' definition (how can the category be recognized ?) but the 'value' definition (what is the meaning or function of the category?) is, also with phenotypes, a very difficult and elusive enterprise (cf HALLIDAY 1988). One only has to think here of the many controversies in modem lin-

- 284-

Leuvense Bijdragen 87(1998)

guistics over the function of fonnally well-established categories such as, for instance, Subject. Cryptotypes are covert grammatical categories. They constitute submerged meanings shown only as an influence (WHORF 1956:105), or to use the metaphor from chemistry which Whorf introduced as a semi-technical term, as a reactance (1956:89). A 'reactance' is the interaction between the meaning of the cryptotype and the meaning of another system in the grammar such as voice, causation, tense, aspect, and nominalization. One immediately notices the basic similarity with Gleason's 'agnation' thinking. Like 'agnates', 'reactances' serve to identify distinct categories that might otherwise remain undetected. Cryptotypical categories are difficult to define both in terms of their realization and their value. Defining them from the realization end requires working carefully through all the possible reactances. The 'realization' definition will always be 'configurative' (1956:80) - but rarely exhaustive at the present stage of research. Importantly, Whorf stressed that the configurative realization of the category provides the basis for a truly meaningful characterization of the value of the category. However, the hazards and difficulties of such interpretative work are obvious. He referred in this context to the "dummy category" of the intransitive used in traditional English grammar, and suggested that a more significant characterization could be arrived at, if the intransitive in English was investigated as the cryptotypical category it really is. It is part of the meaning of the intransitive that it cannot, for instance, be pasivized or causativized (/ went, */ am being gone, *They went me). In section 1.2.1, we will see how, for instance, Halliday's processparticipant roles are such 'fleshed out' cryptotypical categories whose proposed semantic values generalize over whole agnation paradigms. But before we move on, we should first assess the general theoretical relevance of thus involving paradigmatic variants in the defmition of linguistic categories.

1.1.3. Enation and agnation: their relevance to conceptions of the language system and linguistic categories Gleason's views on 'agnation' and 'enation' relations between constructions also imply a specific conception of the language system and linguistic categorization - as non-mainstream and controversial as his methodology. Methodological and theoretical questions are, of course, intrinsically related. A sound methodology should not be a bag of tricks, unrelated to fundamental properties of the language system. Rather, any systematic way of interrogating and modelling language use implies fundamental theoretical claims

- 285-

K.DAVIDSE

, I I

..~. • r.

l~

...

t#1

.~;~~

,

'

about the language system In what follows, I will attempt to explicate the consequences of agnation-enation methodology for one's conception of the language system, by a comparison with LANGACKER'S (1990) 'usage-based' model of language. In the Langackerian model, the 'internal' structure of a category is defined by the categorizing relation of 'instantiation' and linked up with 'schematicity'. Schematicity is Langacker's notion to characterize the instantiation relations between 'higher-level' and 'lower-level' grammatical structures, and between symbolic grammatical constructions and their instantiating expressions (LANGACKER 1990: 15fi). More or less 'schematic' means both more or less 'general' and more or less 'abstract'. Thus, a higherlevel schema in the grammar generalizes over a number of lower-level grammatical constructions. A high-level schema is also the most abstract, in that, although a fully functional structure in its own right, it contains the smallest number of semantic specifications. Grammatical constructions are thus analyzed as complex categories, which can be represented by a schematic network. Such a schematic network captures the categorizing relationships between the most general and abstract schemata and the more specific, structurally-diverse constructions implementing them The schematic network can also model the categorizing relationships between lower-level structures and their instantiating expressions. Langacker further distinguishes the possible categorizing relationships obtaining between the members of a prototype as either: (1) instantiation, (2) extension, or (3) mutual similarity. He characterizes these categorizing relationships as follows. The instantiation relation amounts to a relation of specialization, i.e. including more precise and more detailed semantic information. "Extension ... implies some conflict in specifications between the basic and extended values". Mutual similarity "differs from extension by lacking directionality". To make the link to Gleason, we can note that 'enation' corresponds to co-instantiation between actual linguistic units in Langacker's model. Langacker allows for some 'leeway' in the instantiation relation, in that it need not be complete, but may involve 'extension'. We can think here of Gleason's gradable concept of 'partial enation'. However, Langacker's model does not reach beyond partial enation - his schematic networks do not 'net in' agnates. The concomitant view of the linguistic system is one inhabited by distinct categories whose internal structure is captured by rich and differentiated networks, but whose external relations to each other seem of secondary importance.

- 286-

Leuvense Bijdragen 87(1998)

By giving mutually dependent definitions to the enation and agnation relations, Gleason (re-)values the inter-categorial areas in the languag e system : they play a role in the very constitution of linguistic categories and, presumably, in the emergence of new ones. The inter-categorial areas are as rich and differentiated as the intra-categorial ones. In the next section (1.2.1), we will focus on the question how to picture these inter-categorial areas: are they the same between categories of a different nature such as declara tive interrogative versus transitive-intransitive, and what part does structur al coding play in them?

1.2. Agnates and construction types 1.2.1. The Hallidayan tradition: a metafunctional approach to agnates and construction types

Agnation methodology and the principle of cryptotypical categories were espoused keenly by M.A.K. Halliday. In elaborating a functional gramm ar as expressive of ideational and social meanings, he has always been interest ed in the - interrelated - choices offered by the language system - not in how structures can be formally transmuted into each other, that is, in agnatio n rather than transformatioD. Both his descriptive heuristic and the formali zation of these descriptions have always been strongly paradigmatic: to identify and interpret linguistic categories, he has consistently used 'agnates' and to represent the relation of choice between linguistic categories, he has used system networks. Yet, he has never theorized much about the relation between agnation and his functional categories on the one hand, and that between agnation and system networks on the other. In this section, I will offer some personal reflections on these two questions. Like Gleason, Halliday conceives of the language system as not a mere inventory of structures, but as a network intrinsically defmed by the relations between those structures. Likewise, he adheres to the principle that it is primarily via the paradigmatic axis that one can 'break into' the languag e system. Grammatical paradigms display proportionality (symbolized by:: ). Such proportional relations are, of course, defined by agnation and enation . This can be illustrated by the following paradigm of constructions (HALLIDAY 1994 :xxxii), in which the horizontal relations represe nt agnation and the vertical ones enation. (9) a This teapot the duke gave to myaWlt. (10) a That I told you before.

(9) b The duke gave my aWlt this teapot. (10) b I told you that before.

- 287-

:.

i

K.DAVIDSE

Now, as pointed out by Halliday, examples such as (9)-(10) are proportional in a certain respect. The respect in which they are proportional leads to a linguistic generalization, which is captured by linguistic categories such as, in this case, 'marked Theme/unmarked Theme'. Categories such as these were then ordered by Halliday in terms of dependency: which categories are presuppposed by which? For instance, in order for an English clause to open up the choice 'active/passive', it should be transitive. Such a dependency can be represented by 'system' notation, with an 'entry condition' (transitive clause), opening up the 'system' VOICE, which, in English, consists of the two choices active and passive. active

""'"ti~ ,""" -----~~~~:-----:> [

. passIve

Figure 1 :

the system of voice

In working out these dependencies for clause grammar, he observed (1969) that they clustered, by and large, in three big components, which displayed many internal dependencies within the components, but few relations bridging between the components. He then put forward the claim that these discrete networks encoded different types of abstract functional meaning, viz. textual, interpersonal and ideational meanings. These different domains in the language system thus appeared to be 'meta-functionally' motivated and he therefore referred to them as metafunctional components, or simply, metafunctions. The ideational organization of the clause is concerned with the representation of our experience of the world. Its categories capture various process-participant relations such as Actor-material process-Goal, and Senser-mental process-Phenomenon. The interpersonal organization of the clause construes a situation type as grounded vis-a-vis the speech exchange between speaker and hearer. This layer of organization is constituted by the categories of Subject-FinitelPredicator-Complement(s)-Adjunct. The textual organization of the clause contextualizes its message in the larger textual and situational context. Attempts to capture this type of organization have led to concepts such as Theme-Rheme, and Given-New. While the exact elaboration of these three discrete networks has undergone certain changes in Halliday's own work and that of his associates (e.g. MATTHIESSEN 1995), their basic outlines have remained fairly constant. If we consider these networks more closely in terms of what aspect from the agnationlenation intersection they categorize, we fmd that the text-

- 288-

Leuvense Bijdragen 87(1998)

ual and interpersonal networks basically model 'agnate' structures, whereas the ideational network sets out 'non-enate' configurations. For instance, with interpersonal categories such as declarative-interrogative, the interrogative is a systematic agnate of the declarative. In contrast, with ideational categories such as intransitive-transitive, the transitive is not a systematic agnate of the intransitive, and the former shares hardly any agnates with the latter. They are, therefore, non-enate configurations. There is an important contrast here, calling for further clarification. In a textual network, the basic choices represented, such as unmarked versus marked Theme and various possibilities of Given-New distribution can, in principle, be applied to one syntagm with identical lexical units. That is, the network models (a specific interpretation of) agnation relations. For instance, a clause in which the Theme 'matter' is explicitly marked by as for, is an agnate of the corresponding clause with unmarked Theme. unmarked theme (Theme/Subject)

-7 major clause

[ theme matter [ ('as to/as fori ... + nom Gp) marked theme as non-Subject

Figure 2: (partial) theme network (based on MATIH1ESSEN 1995:540) As correctly pointed out by G6MEZ-GoNzALEZ (1994), Halliday's formal and semantic definitions of textual categories such as Theme and marked Theme reflect his specific interpretation of agnation paradigms such as the following: (II) a Doctor Fischer had stolen my death. (unmarked Theme: Subject) (II) b As for Doctor Fischer, he had stolen my death. (marked Theme : Theme matter) (II) c As for my death, Doctor Fischer had stolen it. (marked Theme: Theme matter) (11) d My death Doctor Fischer had stolen. (marked Theme: non-Subject)

We will not look further here into the specific arguments given by Halliday for these textual categories, nor compare them with alternative generalizations about textual organization such as, for instance, found in the work of the Prague School. What is important to our present concern is to note the essential link between terms in the system network and their formal coding means. The latter involve variation in constituent order, intonation and determiners, i.e. - 289-

K.DAVIDSE

what are generally recognized to be fonnal markers of textual, or 'pragmatic' , categories. Other coding resources, such as case marking and SubjectFinite concord, are simply not involved. This means that, in working out proportionality relations, the formal codings are, apart from the typical textual variation, held constant. This entails further that, from the formal end, the inter-categorial areas between, for instance, clauses with marked versus unmarked Theme, are coded by just a few contrasting structural elements, which are superimposed on a large substratum of shared coding means. It seems to me that this is the typical formal coding of the agnation relation. From the 'coding' end, categories such as clause with marked versus unmarked Theme are clearly closely related to each other in the linguistic system, sharing as they do most of their formal realization. As we will see, a similar picture emerges for interpersonal clausal categories. Interpersonal clausal organization is, in Halliday's description, concerned with fundamental choices relating to the nature of the interpersonal exchange performed by the utterance. For instance, whereas interrogatives and declaratives are concerned with the exchange of 'knowables', imperatives exchange 'desirables'. Networks modelling these choices are also agnation networks, dealing with clausal variants that are, by and large, applicable to anyone clausal syntagm. Take, for instance, the following fragment of the mood network. declarative 'M;~~

[ wh-interrogative

[

major clause

interrogative ---?

[ polar

Imperative

Figure 3:

(partial) mood network (based on MATTHIESSEN 1995:392)

It can be illustrated by agnation paradigms such as (12) a (12) b (12) c (12) d

We should tell Doctor Fischer the truth. (declarative) What should we tell Doctor Fischer? (wh-interrogative) Should we tell Doctor Fischer the truth ? (polar interrogative) Tell Doctor Fischer the truth. (imperative)

As observed by HALLIDAY (1994), the fonnal marking of these agnates clusters in English around the elements 'Subject' and 'Finite verb'. The distinction between indicative and imperative is marked by the presence or absence of Subject and Finite, that between indicative and interrogative by the relative order of Subject and Finite. So, here too, formal realization of

- 290-

Leuvense Bijdragen 87(1998)

agnate constructions resides in just a few contrasts on a substra tum of shared coding means. Let us look in more detail at the categories 'declara tive' and 'interro gative' and summarize a few well-known facts about their gramm atical coding (cf QrnRK et al 1972, MARTIN 1992, MATIHIESS EN 1995). Declaratives have non-inverted Subject-Finite order and are spoken on a falling tone. These coding means go into expressing the declara tive's function of 'asserting' information about a situation. Polar interrogatives, in contrast, have inverted Finite-Subject order and are spoken on a rising tone. These formal means express the structur e's function of 'un-ass erting' information in the sense that it inquires whethe r or not a situation occurred. Wh-interrogativ es, then, have inverted Subject-Finite order, except when the Subject is the whitem, and are spoken on a falling tone. Their semantic value reflects this 'hybrid' coding, since they 'assert' a lot of information about a specific situation and inquire only about one elemen t of that situation. At the 'coding ' interface of the language system, the 'inter-c ategori al' area betwee n declarative and polar interrogative thus appears as defined by a few formal contrasts and a lot of shared coding. Wh-interrogatives occupy a space which intersects, to a certain extent, with formal and semantic features from declarative and polar interrogative. Categories such as declara tive and the two types of interrogative do not appear here as purely internal ly differentiated and externally unrelated. Rather, at the coding interface of the linguistic system, their form-meaning components appear as a networ k which defines both their intra- and inter-categorial relations. We can conclud e that interpersonal agnate constructions, just like textual ones, appear, when considered as coded categories, as largely overlapping sets of form-meaning correlates, i.e. as intersecting areas of the coding network. It may be instructive at this point to refer to the views of the languag e system which THIBAULT (1997:Ch.3) ascribes to Saussure. He claims that Saussu re's well-known definition of the language system as an 'opposi tional system of pure values' (referred to as ltmgue 1), is comple mented in the Cours with a second theoretical perspective on the language system, langue 2. This second definition ''focuses on the regular and typical ways i:D. which associative and syntagmatic solidarities are formed in the language system " (THIBA ULT 1997:64). These two perspectives on the language system can be applied to the agnate categories of declarative and interrogative, which we have just considered. Langue 1 constitutes the more abstract perspective : it is concern ed with the distribution of conceptual distinctions among the terms in a given language system, where 'terms' are conceiv ed of as more abstract units than

- 291 -

_I'

?,

K.DAVIDSE

'signs', referring neither to form nor to substance. From this perspective, 'declarative' and 'interrogative' appear as abstract terms dermed by purely differential relations. However, when the semiotic form-meaning couplings of langue 2 are brought into the picture, then declarative and interrogative clause types appear as sign types motivated in a positive way. Formally, they are defined by overlapping sets of coding features. They thus construe specific semantic values, in which, likewise, differences in a few vectors are superimposed on a substratum of shared functional meaning. We can now wind up this discussion of interpersonal and textual clausal categories. The distinct categories traditionally considered in these domains, such as clauses with marked versus unmarked Theme, and declarative versus interrogative clauses, are agnate categories. In these areas of the grammar, paradigmatic variants are the distinct categories to be elucidated formally and functionally. If we then turn to the ideational domain, we rmd a different relation between categorization and agnation. The most general categories, for instance intransitive, transitive and ditransitive clause types, are not agnate but non-enate with regard to each other. This means two things. Firstly, they are not 'agnate' in the obvious sense that intransitive, transitive and ditransitive syntagms involve distinct lexical items; in particular, they are associated with distinct verb classes. Secondly, they are 'non-enate' in the sense that they do not share whole clusters of agnates. For instance, as pointed out by Whorf (section 1.1.2), intransitives do not have the passive and causative agnates associated with the transitives, while transitives lack all the variation which is caused in the ditransitive paradigm by changing the relative order of the two non-agentive constituents (see section 2.1). The 'inter-categorial' space between an intransitive and a ditransitive is much larger, so to speak, than that between, for instance, a declarative and an interrogative. Categories such as intransitive-ditransitive are divided from each other by their distinct sets of agnates. As is well-known from valency studies, these two aspects are related : there is a link between specific classes of verb meaning and the specific construction paradigm which these verbs allow. Where do agnates come in, then, in the ideational domain, if they do not constitute the primary categorizations? Let us consider, for this purpose, two specific ideational categories distinguished by HALLIDAY (1968:196), viz. the 'goal-directed' and the 'descriptive' type of transitive. The former, e.g. John hit the ball, is analyzed by Halliday in terms of the process-participant configuration Actor-process-Goal, the latter, e.g. John marched the soldiers, as Initiator-process-Actor.

- 292-

Leuvense Bijdragen 87(1998)

intransitive (Actor - process) e.g. John is running. major clause --7

[

transitive

--7

goal-directed (Actor - process - Goal) e.g. John hit the bal/.

[

descriptive (Initiator - process - Actor) e.g. John marched the prisoners.

Figure 4 :

(partial) transitivity network

In his argumentation for setting up these distinct role configurations, HALUDA Y (1968: 196) brings in agnates such as non-cau sative versus causative fonDS (analytical and synthetic) and clefts with pro-verb do. The ideation al agnation sets he thus considers can be illustrated as follows: Paradig m!: (13) a John marched the prisoners. (13) b John made the prisoners march. (13) c The prisoners marched. (13) d What the prisoners did was march. (13) e What John made the prisoners do was march.

Paradigm II: (14) a John hit the ball. (l4) b *John made the ball hit. (14) c *The ball hit. (l4) d *What the ball did was hit. (14) e *What John did the ball do was hit.

These distinct agnation paradigms are used in the fIrst place to identifY (13a) and (14a) as distinct ideational configurations. Secondly, they are exploite d as a heuristic to interpret the semantic difference between (13a) and (14a) : the agnates in Paradigm I show the prisoners to be 'doers', or Actors, whereas the corresponding agnates in Paradigm IT show the books to be 'under. goers', or Goals. We can now be clearer about the role played by agnation in ideational categorization. Ideational system networks such as Figure 4 represe nt nonenate confIgurations, but 'behind ' each non-enate confIguration, there is a specifIc agnation paradigm, on the basis of which that particular confIgu ration was identifIed in the fIrst place and then interpreted semantically. The general categories in the ideational domain thus formulate general izations about whole agnation paradigms. Agnation paradigms such as (13a-e) and (14a-e) are presumed to share crucial semantic features pertaining to, for instance, different types of agenthood and patienthood. . In a second movement, however, that agnation paradigm itself can and should - be subjected to more detailed analysis : the individual agnates within it should be further characterized. That is to say, a description should

- 293-

, .. ;;:

::>

,'~

K.DAVIDSE

be elaborated of what distinguishes them, beyond the general characteristics which they share. For instance, all ditransitive agnates share general characteristics which can be reflected by a general semantic schema and role configuration, which, I (1996a, 1996b) have proposed, is that of causation by an Agent of a 'possessive' implication between Dative and Patient. However, at a more specific level of analysis, the grammatico-semantic characteristics of the individual ditransitive agnates have to be accounted for. At this level, I developed a description distinguishing Dative-oriented, or 'benefiter', from Patient-oriented, or 'transfer', construals, as well as 'donatory' from'destinatory' ones. (For further discussion of these, I refer to previous publications (DAVIDSE 1996a, 1996b) as well as to the second section of this article.) Given the primary discriminating role accorded to construction paradigms, what importance is attributed to structural coding means in an agnation-based approach to ideational role configurations? In agnation-based methodology, paradigmatic generalizations logically precede structural interpretation, whereas in a syntagm-based aproach, paradigmatic considerations are not brought to bear systematically on the interpretation of structure. Let us illustrate this with existing interpretations of ditransitivity. As an example of the syntagm-based approach, consider PINKER (1989), who analyzes all following three examples in terms of the thematic role configuration 'Agent causes Theme to move to Direction'. (15) They drove the car to Chicago. (16) They threw the ball to Mary. (17) They donated money to Unicef.

This configuration does not account for related (18) They threw Mary the ball

which, following Pinker (ibid.), should be analyzed as 'Agent causes Possessor to have Theme'. In contrast, in an agnation-based approach such as HALUDAY's (1994: 145), (19a) and (19b) should both be analyzed as an Actor-Goal-Beneficiary constellation, while (20a) should be analyzed as Actor-Goal-Place:motion towards. (19) a She sent her best wishes to John. (19) b She sent John her best wishes. (20) a She drove the car to LA. (20) b ·She drove LA the car.

In the syntagm-based approach, the sets of construals over which semantic, or 'case', roles generalize include identical-looking syntagms and exclude - 294-

Leuvense Bijdragen 87(1998)

syntagms with a structural coding difference. It is thus a very abstract structure, viz. NP-VP-NP-to-NP in the case of (15)-(16)-(17), which detennines the class membership of an ideational configuration. Generalizations about the semantics of the configuration are primarily based on the very schematic properties of this structure, such as constituent order and presence of prepositions. The presence of preposition to in (15), (16) and (17), for instance, has, no doubt, been an important factor in the analysis of to Chicago, to Mary and to Unicef as Direction and for the concomitant semantic generalization over all the VPs as members of the 'move' class. In an agnation-based approach, by contrast, the apparent identity of syntagms such as (19a) and (20a) is treated as a case offalse enation, which is disambiguated by their distinct agnates (19b) and (20b). The sets of constructions over which participant roles generalize in this approach are defined by agnation. Each of these agnation sets is characterized by a basic role configuration and a different verb class (see below). Thus, ditransitive verbs, which contain the semantic component 'make someone the possessor of something', constitute a different verb class from verbs which mean 'act on an object so that it is moved to a specific direction'. The verb send in example (19) belongs to the former class, while drive in (20) belongs to the latter. The interpretation of structure comes in after this agnation-based disambiguation. This raises the question of the level of schematicity at which structure is interpreted in this approach. There are two possible answers to this question. Either one could hold that the structural meaning in examples like (15)-(17) is always the same - and thus has to be interpreted at a highly schematic level - and that the distinct ideational meaning of these configurations comes entirely from the verbs, as members of distinct verb classes. Or one could hold that structural markers have to be interpreted at a more specific level, i.e. as encoding meaning specific to each agnation paradigm. Schematic structural markers are then viewed as being 'instantiated' in specific configurations and construing 'more specialized' meaning in accordance with the distinct agnation paradigm. With regard to a very schematic structure such as NP-VP-NP-to-NP, I am thinking of the following sorts of structural differences which it may acquire within specific construction paradigms. There may, for instance, be different patterns of inherency and directionality internal to the structure. Consider, for instance, examples (21)-(22), the first of which is 'transitive' and the second 'ergative' (as defined in DAVIDSE 1992). (21) He posted a present to my wife. (22) A latent carrier may pass the disease to descendants.

- 295-

;;

..

)

;..

........

K.DAVIDSE

The semantic difference between (21) and (22) has commonly been explained purely as a consequence of the lexical differences between 'transitive' and 'ergative' verbs. However, my position is that there are constructional differences between transitive and ergative construals. The 'transitive' clause (21) expresses a unilaterally 'effective' relation: the process is targeted by the Agent onto the two non-agentive participants, which 'undergo' the process in a fully 'passive' manner. That is, their participation in the process is wholly determined by the Agent's action. The Patient-Dative unit has no structural 'independence', which is borne out by the fact that this unit cannot receive independent circumstantial modification. We cannot use an utterance like He posted a present to my wife on Monday to express that the present reached her on Monday. The 'unilateral' relation linking Agent and non-agentive participants is represented by a simple arrow in the following representation: (21)' Ag - proc

I

II

I~

.. :l!

........ ~

Ic

~

Pa to Dat

The 'ergative' clause (22) expresses a cyclical form of causation, involving two processual cycles : the 'passing' of the disease to descendants, and the instigation of this transfer by the latent carrier. Structurally, the Patientprocess-Dative unit has a certain independence, which is shown by the fact that not only the Agent-process relation, but also the Patient-process relation may serve as 'elaboration site' (LANGACKER 1991:205-206) for circumstantial modification. Thus, in A latent carrier may unwittingly pass the disease to descendants in the next generation, unwittingly modifies the Agent-process relation, whereas in the next generation modifies the Patientprocess-Dative unit. In (22)' the semi-independent participation of the Patient in the process is represented by a second, separate, arrow, and the relative structural independence of the whole Patient-process-Dative unit is indicated by the brackets surrounding this unit : (22)' Ag ~ (proc {- Pa to Dat)

This is just one example of a 'lower-level' structural difference, but the idea is that the relations between NPs and VPs in structures belonging to distinct agnation paradigms are characterized by such systematic differences in terms of 'directionality' and 'autonomy'. I favour this approach to structure, precisely because it can account for structural meaning both at highly schematic and more specific levels. At a very schematic level, it allows one to capture the obvious difference between, say, a transitive versus a ditransitive structure. At more specific levels, fmergrained differences between different types of transitives and ditransitives can be accounted for in terms of such features as directionality of relations

- 296-

Leuvense Bijdragen 87(1998)

between NPs and vps, relative autonomy of NP-VP units, and inheren cy/ optionality of NPS. Moreover, this view on structure can be harmonized with that on verb classes which I will propose (see section 2.2). As we will see, semantic features relating to directionality and inherency are also part of the semantic make-up of the verb classes carved out by agnation methodology. The aim of this section was to shed more light on the relation between agnationlenation methodology and the metafunctionally diversified categories of clause grammar found in the Hallidayan tradition. Summarizing, we can say that textual and interpersonal clausal categories, such as clause with marked versus unmarked Theme, and declarative versus interrogative, are agnate categories. General ideational categories, such as intransitive, transitive and ditransitive, are non-enate, but develop into partially non-ena te and partially enate, as the description is refmed in delicacy; both the general categorial cuts and the more specific ones are based on the differen ces between the agnation paradigms behind these distinct ideational configu rations. It is these 'differences' that, in a Hallidayan approach, one attempt s to capture by categories such as Initiator, Actor and Goal, which reflect different grammatico-semantic forms of agenthood and patienthood, variatio ns on the 'inherent voice' (HALLIDAY 1968:198) by which nominal constitu ents are related to the verbal unit. Ideational roles are thus categories which intrinsically generalize over their specific agnation paradigms, i.e. over all the other manifestations of that specific process-participant configuration. In contrast, interpersonal categories (Subject, Finite, Predicator, Comple ment, Adjunct) capture the syntactic relations within one specific svntagm. Textual categories such as Theme-Rheme, Given-New, capture the 'informational nexi' (DANES 1974:109) of the contextualized svntagm.

) )

..

>I

1.2.2. Vande n Eynde : verb-general and verb-specific variants of constru ctions In the previous section, we have seen how, in the Hallidayan tradition, agnationlenation relations between constructions are described in a metafun ctionally diversified way. Textual and interpersonal construction types are agnate, while general ideational configurations are non-enate with regard to each other. There are some obvious parallels here with the methodologica l proposals made by VAN DEN EYNDE (1995) for valency research. As a starting point, he takes the tenet that the relations between different constru ctions of the same verb are an essential constituent of the syntactico-semantic network. Regarding these paradigmatic variants, he distinguishes between verbgenera l and verb-specific ones. Examples of verb-general variants are anteposition of non-subjects and interrogative inversion, which he labels dis-

- 297-

K.DAVIDSE

positives. Verb-specific variants are further divided into reformulations, which, while being verb-specific, have some regularity, like the passive, and linked constructions, which lack that form of regularity. The dispositives seem to correspond roughly to Halliday's textual and interpersonal agnates, the reformulations and linked constructions to the ideational agnates. Also from a semantic point of view, VAN DEN EYNDE (1995:119) radically inserts himself in the paradigmatically-oriented approach to constructions with the following tenet : "The meaning determining unit is not the construction itself, but the network it is part of'. He thus uses reformulations and linked constructions to identify distinct constructions, which often involves "disambiguating" (ibid) apparently identical syntagms. For instance, he shows (1995:118) a syntagm such as (23) Je l'ai double(e)

to be four-way ambiguous. A first construction applicable to that syntagm, in which the Patient role could be filled out by Ie prix, has the following set of reformulations and linked constructions behind it