The contrastive component: Frames and semantic fields

We can think of the linguistic frame as a semantic (microuniverse) which on the one hand reflects the regularity of typical contexts, and on the other reflects the structured nature of the lexicon, in other words, the contrastive aspect of lexical meaning. We know that the lexicon is not a list but an organized set of relations and connections that delimit semantic subsets: father, mother, brother, sister, uncle, brother-in-law, etc. form the microuniverse of family relations; buy, sell, spend, pay belong to the schema of the commercial event; day, night, afternoon, evening to the system of temporal division. The meaning of single terms cannot be defined independently of the set of relations that constitutes the frame itself. Single words are understood in terms of the structural relations between all the words that make up the overall frame of that particular lexical domain. This position is cer­tainly in tune with the fundamental inspiration of structuralism, which is the semantic theory that has primarily emphasized the contrastive aspect in semantic analysis, and which began with the concept of positional value introduced by Saussure. By distinguishing between signification and value, Saussure wanted to strip the denotative dimension of meaning of its importance and to position lin­guistic meaning within defined intralinguistic relations. As I argued in chapter I, I do not believe meaning can be reduced entirely to the relative positional value of terms, because the sum of intralinguistic relations is not in itself sufficient to establish the meaning of a term if it has not already been interpreted by other means. Whatever the nature of this process of interpretation, it always involves an "overflowing" of the linguistic. Or, to put it in more semiotic terms, it involves triangulation with another kind of element.

Although meaning cannot be only contrastive, it certainly is also contrastive, and the specification of positional value is an integral part of an adequate seman­tic representation. From this point of view, the frame could be defined as a stable structured system of connections between positional values. Indeed, this is the intui­tion underlying early works on the theory of semantic fields in the twenties and thirties (Ipsen 1924, Trier 1934, Porzig 1934) right through to more recent times (Coseriu 1974, Lehrer 1974, Lehrer and Kittay 1992). Semantic fields and frames certainly have a great deal in common: a similar non-atomistic vision of meaning, which, to borrow an expression from Lehrer and Kittay (1992: 12), is definable as a local holism, that substitutes the global holism of Davidson and Quine with a structured description of the way in which the meaning of a term is delimited by other terms in the same conceptual domain. In both the theory of semantic fields and that of frames, understanding the meaning of a term means understanding the structure in which that particular term plays a role. The main difference be­tween them, however, lies in the way this structure is interpreted.

According to semantic field theorists, this structure is entirely lexical and not conceptual, that is, it is constituted by the existence of other words (cf. Trier 1934). The meaning ol each term is thus defined by the set of lexical relations with other terms, and (he entile vocabulary ol a given language is composed ol closed classes

of semantically interdefined terms. In a semantic theory based on the idea of cognitive schemata, terms are connected in relating to the same unitary and co­herent schematization of experience or knowledge, and for this reason the under­standing of individual terms is at the same time the understanding of their set.¹⁵ One could say that semantic fields are a theory of the relations between words, while frames are a theory of conceptual organization and only through this are they a theory of the lexical units that represent concepts. In a frame theory, the meanings of words are not directly related to each other, as in lexical fields; mean­ings are linked only through the mediation of their shared membership of the same cognitive schema, onto which they are projected according to particular per­spectives, and with a focus on different dimensions.

As a consequence, there are also differences in the ways that understanding processes are regulated: to understand the meaning of a term one must first un­derstand the frame underlying the concept expressed by that term. An example of the difference between the two approaches is given by Fillmore and Atkins (1992: 77) regarding nouns indicating the days of the week. From the perspective of semantic fields, the words Monday, Tuesday, Wednesday, etc. constitute a closed class of terms linked to each other by a "successive" relation that defines a cycle. The terms forming the set are interdefinable on the basis of their positions in the cycle, established by the relation of succession and the "part of" relation which connects them all to the term week, the name of the whole cycle. A description of the same terms based on the idea of the frame, on the other hand, would empha­size their position in the entire system of terms relating to the calendar. To under­stand the names of the days of the week and the other terms connected to them, it is necessary to refer to a single interpretative frame relating to the natural cycle created by the apparent daily movement of the sun, the standard procedures by which we establish the end of a daily cycle and the start of the next one, the conventional cycle of seven days and its conventional beginning, and the practice of our culture to assign different parts of the weekly cycle to work and to non-work. An implicit knowledge of this particular physical and social organization of our world provides us with the conceptual base for the understanding of a vast amount of lexical material, not just the days of the week. Using the language of Gestalt theory, Fillmore suggests considering the background of practices and knowledge that constitutes the frame underlying the vocabulary of a particular area as the background in relation to which each single term represents the figure.

Two features differentiate the schematic model from the field model: first, the more markedly encyclopedic nature of the frame of reference adopted, which takes in a broad and complex description of our understanding and experience of time; secondly, the fact that it is conceptual and not only lexical. This second point marks the most significant difference, because it relates to two different theories of meaning, intralinguistically defined in one case, conceptually in the other. A schematic semantics like rhe one outlined thus far is entirely cognitive: lexical meaning cannoi be defined independently ol our conceptual structures and has more to do with this conceptual substratum than with the relations between

Regularity and Context

lexical terms. This is a position that the whole of cognitive linguistics shares today, and it is well illusttated by Fillmore's position. He considers as a paradigmatic case the term hypotenuse. If meaning were only a function of lexical relations with adjacent terms, the wotd hypotenuse should have two different meanings, one in languages like English which have only this wotd, and the othet in languages like German and Italian that lexically conttast hypotenuse with kathete (in German, or cateto, in Italian, referring to the other two sides of a right triangle).

The word field theory, taken literally, would predict that the English word hy­potenuse and its German homograph have different meanings, because only in the German case do mere exist relevant lexical partners for the "mutual defining" to have its effect. It would further be predicted that the German school child who knows the word hypotenuse but unlike his class mates has not learned yet the word kathete does not have the same meaning for the former word that his com­panions do. This, of course, cannot be believed. (Fillmore 1985: 229)

In a schematic and conceptual semantics, on the other hand, the meaning of each of these terms is directly connected to the background frame, which in this case is given by knowledge of the chatacteristics of a right-angled triangle. It is on the basis of this knowledge that we fix the meaning of hypotenuse, which is thus in­dependent of the existence of a specific lexicalization for the other two sides of the triangle.

In this way, meaning is definable, as has alteady been seen, not only through "negative" differential descriptions, what a term is not, but also through positive qualifications of what a term is, qualifications that cannot be entirely reduced to inttalinguistic telations. It thus proves possible to produce a tepresentation of in­dividual lexical units, a possibility that is automatically excluded from field the­ory, as from any strictly structuralist representation. Naturally, the two dimen­sions (the positive and the differential) are interwoven and can take on different degrees of importance depending on the case; if in certain semantic configura­tions the relation between terms is determinant, there are, however, also cases in which the understanding of a wotd is independent of the knowledge of all the other connected paradigmatic terms. This seems to be true, for example, for terms that relate to phenomenologically founded experiences: Fillmore uses the example of scratch, but think also of yawn, headache, nausea, tickle, etc. In all these cases it is the phenomenological experience of the bodily sensation concerned which allows us to understand the meaning of the word, although other related words may contribute to its lexical delimitation.

What this means, and this is the most important theoretical consequence, is that the teference schemata are not always supplied by language; although in some cases they may be created by language, in many othets they derive from non-linguistic experience, and manifest themselves as having an existence that is autonomous and independent from linguistic structure.¹' This assumption marks a radical move away from structuralist positions. To conclude, a schematically based semantics comprehends the contrastivc aspects ol meaning, (bus permitting

the representation of the positional value of terms;¹⁷ however, it is not limited just to the differential dimension, in that it always connects lexical meaning to an underlying regularity which is not linguistic but experiential and cognitive in a broad sense.

8.}.}. The narrative component of lexical meaning

A schematic representation like the one that we have outlined so far can also be distinguished from the theory of semantic fields by its ability to explicate the tem­porality within the semantics of many terms. Fields are static representations that describe a specific set of structural relations lacking in internal dynamism. Now in many cases, even the semantic relation between lexical units requires reference to the "history of events" that created that relation. For example, the configura­tion of the commercial exchange that we analyzed in the previous section requires in reality a dynamic schematization that corresponds to a fully fledged narrative (see diagram 18), as indeed is suggested by the schema proposed by Fillmore (1976b: 105). The order of the boxes from left to right indicates temporal transfor­mations, while vertical alignment indicates simultaneity. A schema of this kind is nothing other than the depiction of a minimal narrative program¹⁸ that involves an initial state, defined in this case by a relation of conjunction (that is, of pos­session), in which two subjects (A and D) are respectively in possession of objects С and B. Following an agreement or contract between the two subjects, action X ensues, in which the two subjects exchange С and B. When this action has been completed, the final state, Y, is achieved, in which the relations of possession are opposite to those of the initial state. In this schema, the meaning is no longer defined as the result of a series of paradigmatic oppositions but is inscribed within an openly syntagmatic dimension.

The narrative dimension is particularly evident in the semantics of predi­cates;¹⁹ in fact, all verbs relate to an underlying schema of actions that, through such positional roles as subject or object, already prefigure a minimal narrative structure. All of this was evident from the first formulations of case grammar, in that the cases themselves can be read as actant roles around which the narrativity of the predicate is articulated.²⁰ Of course one could object that not all terms in the lexicon are affected to the same degree by a narrative dimension, nor do they all need to refer to temporal succession to the same extent. From this and other points of view, the lexicon cannot be considered a homogenous system, as we will see more clearly in chapter 10. An initial difference at this level can undoubtedly be identified between verbs and nouns: if verbs express a relation between entities and seem to relate immediately to a narrative microstructure, this does not seem to hold true for nouns. Within nominal semantics, natural kinds seem to be less "narrativizeable" than terms that refer to culturally defined concepts whose mean­ing could not be literally explained without a history of the events that have formed them. There are many examples of this kind: terms like alimony or ransom

commercial event

- possess А С

— agreement -

A, D X > Y

- possess D В

- X —,

give ACD

give DBA

possess AB

possess DC

Diagram 18

relate to complex narrative scenes that can be seen, following the suggestion of Eco (1979) as genuine micro-intertextual frames.

In Fillmore's analysis, the meaning of alimony, foi example, is projected onto a prototypic scene, or regular context, in which there are two individuals who married at one moment of the temporal sequence and then subsequently divorced and reached an agreement whereby one of the two (typically, at least in a particu­lar age, the man) pays a sum of money to the other. This sum of money is what we call alimony. Whatever format is chosen for the semantic representation of this term, it must take into consideration this complex narrative sequence (and will probably also have to include further information about legal aspects of divorce, possible disagreements, etc.).

The case of ransom is similar. In this case too, understanding of the term involves understanding of a complex narrative sequence in which fiist of all a subject (B) takes away an object of value from another subject (A); subsequently В offets to give it back upon payment by A of a certain sum of money. The third stage is when A gives В the agreed sum of money, which is named the ransom. Note that this narrative sequence is activated and imposed upon the context every time that the tetm is used: if what A pays В is called the ransom, we have the right to assume that В has illegally taken possession of something or someone (C) im­portant to A, that В has offered to return С on the condition that A hands over a certain sum of money in exchange, and that A has agreed to this.²¹ As I have already argued, the lexical selection creates and activates its own context of refer­ence, which in this case is a complex micro-intertextual frame. The tetm thus condenses a narrative configuration that can then be expanded into a whole text. A similar syntagmatic and narrative perspective is to be found within generative

semiotics, in Greimas' analysis (1981) of anger, the meaning of which is broken down into a complex sequence of states and passionate transfotmations that move from expectation to frustration to discontent to aggression.

I believe that a narrative component is present in an enormous number of words, though maybe less explicitly than in the examples outlined above. We have already seen that in the semantics of artifacts one can talk about an implicit func­tional history that refers to the typical sequence of actions envisaged by their function: fill a glass and drink, pick up a pistol and fire, sit down on a chair. The regular associations between functional terms and the respective functional predi­cates {glass/drink, chair/sit, pistol/fire, etc.) constitute a minimal sequence of narra­tive action.

In other cases, even when the semantic definition seems to be based on non­dynamic properties, there is still an implicit narrative temporality that emerges from the possible inferences activated by the term. Take a term like puddle: it can be defined as "muddy pool of rainwater." But the word rainwater already implies a sequence of events where in some moment prior to the existence of the puddle it rained. All of this is part of our knowledge of the meaning of puddle and this determines the possible inferences. Consider the difference between (10) and (11):

10. The street was full of puddles. It had not rained for a long time.

11. The street was crossed by a rivulet. It had not rained for a long time.

Example 10, unlike (11), appears incongruous because it contradicts the most ob­vious expectation connected to puddle: if puddle then prior rain. This naturally does not mean that such a sequence is semantically impossible or "wrong" but only that, by violating the inferential system suggested by the term, a system of 'attention alert" is activated and we expect an explanation of why there are puddles without rain. This does not occur in (11), because rivulet does not presup­pose the same sequence.

One could therefore say, at least in a great many cases, that lexical units are the site of condensation of an underlying narrative structure. This is an idea found in all semiotic thinking, both in the interpretative semiotics of Eco, which is con­cerned more with the processes of the reader's interpretative co-operation, and in the generative semiotics of Greimas. Eco (1979) suggests looking at the sememe as an instruction oriented toward the text; in this perspective, the individual term already contains its potential textual and narrative development, which will be actualized in the moment it is inserted into a particular co-text. In an instructional semantics like the one suggested by Eco, oriented toward textual insertion, and which is inferential by nature, the lexical level and the textual level are closely inter-related. The mechanism of condensation and expansion between the lexical and the textual levels is also stressed by Greimas: a lexeme like fisherman, for ex-ample, is

a very condensed surface construction; it describes a person who possesses a competence limited to a particular 'doing' that is liable to expansion, a 'do­ing' that can cover a vast discursive sequence once it has been explicated; but which maintains, at the same time, and more or less at the same level, its seman­tic character It can occupy, in the two grammars—linguistic and narrative— different actantial positions. (Greimas 1973; my translation)

Fisherman thus carries with it "all the possibilities of its 'doing,'" that is, the be­havior one can expect from that petson; it refers both to a thematic role and to a potential narrative program.

Lexemes are therefore surface condensations of deepet (narrative) configura­tions and in the reciprocal relation of condensation and expansion with wider tex­tual units is manifested that fundamental characteristic of language that Greimas has defined as the "elasticity of discourse." By virtue of this elasticity, a correspon­dence is established on the semantic plane between units of different "sizes" and dimensions that can, however, be recognized as semantically equivalent. Ransom can thus be a narrative sequence, an entire text, or a lexeme: in all these cases the underlying conceptual-narrative structure (the scene, in Fillmore's terms) will be the same. The dialectic expansion/condensation that characterizes the elasticity of discourse can be compared with the relation that links Fillmore's scenes with their surface lexicalization. In fact, in both cases the movement between configurations of greatly different dimensions is possible because we are not only talking about words and relations between words (as was the case in the theory of lexical fields), but about conceptual structures and underlying narratives.

This allows us to deal in a different way with the issue of the "size" of the analytic units. The validity of analysis at a lexical level has often been questioned in the field of semiotics, the view being that only the textual unit is pertinent. But the text versus word opposition vanishes, or rathet, it ceases to be significant, in a schematically based semantics like the one I am trying to delineate. If lexical units are seen as points of condensation, they are virtual texts, to use Eco's expression, and an equivalence is established between surface realizations that may vary con­siderably in theit dimensions.