AN AGENDA FOR TEXT SEMANTICS:

MEANING AND PARAMETRIC ADJUSTMENT

Robert de Beaugrande

To catch the meaning out of the air

Yet to have it inviolably there

— Richard Eberhart, Hardening into Print

1. Projects and parameters in semantics

1.1 Although “semantics” has had a long history, the field remains diffuse and refractory to define, more an agglomeration of problems and proposals than a continuity of specified research methods. No doubt this condition reflects the complex and multifarious nature of “meaning” itself. Human cognition and communication are so omnipresent and adaptive that the possible “meanings of meaning” can seem endlessly diverse. At present the only thing we seem to “know for certain” about meaning is that the “ theories of meaning” propounded thus far are much simpler and tidier than the phenomena involved. At some point, extending and complicating available theories will not be sufficient, and new ones must be developed.

1.2 One way to assess the unity and consistency of semantics would be in terms of major projects. Here are some plausible candidates:

1.2.1 Semantics should explain the nature of meaning.

1.2.2 Semantics should develop a consensual representation for meanings.

1.2.3 Semantics should relate meaning to other human concerns, such as “language” and “knowledge”.

1.2.4 Semantics should inquire how meaning is constituted and used by human beings.

1.2.5 Semantics should inquire how the capacity for being meaningful is acquired and exercised.

Ideally, all these projects could be pursued in conjunction. Semanticists would doubtless be gratified to attain an explanation or a representation of meaning that would also relate meaning to language and knowledge and suggest how meanings are acquired and used. In practice, however, such ideal conjunctions have seldom been realized. More often, one project was singled out, notably the creation of representations for meaning that were not intended to correspond to psychological realities in the acquisition and use of meanings (cf. 1.15).

1.3 Another tactic for assessing unity and consistency would be to envision a set of parameters between complementary criteria that are implied in the design and development of semantic theories, e.g.:

1.3.1 General versus specific. Semantics might seek to explain “the nature of meaning” in the widest sense; or it might examine only the meanings within a specific domain, such as a “special-purpose language” or a “semantic field”.

1.3.2 Deterministic versus non-deterministic. The prevailing view has generally been that meaning is basically determinate except in odd cases of ambiguity or disturbance. A less-favoured view of meaning as basically indeterminate until it is put to some use has been raised in such trends as phenomenology and cognitive science. In yet other trends, such as mysticism, transcendentalism, and deconstruction, meaning is held to be ultimately and irretrievably indeterminate.

1.3.3 Abstract versus concrete. Meaning may be seen as something quite removed from ordinary experience and perception, or as something quite proximate to them. The abstract outlook might cite mathematical or logical terms, while the concrete outlook might cite tangible object-referents in the “real world”.

1.3.4 Objective versus subjective. Meaning would be “objective” if its primary source and determiner is some externally experienced object whose existence and qualities can be publicly constated. Meaning would be “subjective” if its primary source and determiner is some internally experienced construct whose existence and qualities can only be privately constated.

1.3.5 Mentalist versus physicalist. In one traditional view, meaning is an essentially mental entity, something existing or occurring in the mind. In a less traditional view favoured by such trends as empiricism, positivism, and behaviourism, meaning is an aspect either of physically manifest objects and events or else of manifest behavioural responses to them.

1.3.6 Idealized versus realistic. “Idealism” asserts the priority and permanence of the “ideal” meaning existing in a realm independent of particular instantiations or referents, and suspects the “real” of being precariously transitory and accidental. In contrast, “realism” anchors all experience and significance in reality or some reasonably proximate interpretation of it, and suspects the “ideal” of being a product of wishful thinking.

1.3.7 Speculative versus empirical. Semantics can proceed, and has traditionally done so, by constructing theses out of speculations about the nature of meaning, some plausible and conmonsensical, some less so. Or, semantics can attempt to find empirical correlates for theses in the meaningful activities of representative groups of humans.

1.3.8 Static versus dynamic. The view that meaning is a stable and unchanging entity understandably appeals to scholars trying to describe or define it. The view of meaning as a continual flow or process is less appealing but has gained some support in recent years.

1.3.9 Synchronic versus evolutionary. One can accept a meaning as a phenomenon in its current state, or one can seek to explain how a meaning originated and developed, e.g., by pursuing etymologies of words. Prior to the advent of modern (“synchronic”) linguistics, most semantic approaches informally combined the two outlooks.

1.3.10 Knowledge-dependent versus knowledge-independent. Semanticists disagree whether meaning should be treated within a more general theory of knowledge, or as an issue in its own right. This parameter becomes acute when deciding which aspects belong to philosophy and which to psychology.

1.3.11 Language-dependent versus language-independent. Semantics has often had recourse to language for illustrations, but has not established whether language is the paradigm case or indeed the only reliable source of meaning. Today. “semantics” is still reckoned sometimes a part of linguistics and sometimes a part of philosophy.

1.3.12 Language-centred versus participant-centred. Frequently, “meaning” is treated as a event or action performed by words and expressions themselves. But the notion of meaning as an entity intended and generated by the participants in communication has gained ground in recent years.

1.3.13 Denotation versus connotation. Semanticists have customarily argued that the literal meaning of a word (say, its dictionary definition), should be distinguished from experiential or emotional associations it may elicit among particular people.

1.3.14 Componential versus holistic. For purposes of analysis, it may seem convenient to suppose that the meaning can be decomposed into more primitive elements and conversely, that the meaning of a whole (e.g. a sentence) is the sum of the meanings of the part s (e.g. the words}. A holistic approach, in contrast, would maintain that meaning cannot be decomposed without altering it, and that the meaning of the whole can be more than or other than the meanings of the parts.

1.3.15 Linear versus non-linear. Meaning could be “linear” in the sense of being generated from a sequence of items succeeding one another either in time (e.g. speech) or’ in space (e.g. writing). Contrarily, meaning would be “non-linear” when associations are made between items either not appearing in sequence, but, say, in a hierarchy.

1.3.16 Arbitrary versus motivated. A foundational thesis in modern linguistics was that the relation between word and meaning (or “signifier and signified”,. etc.) is arbitrary, i.e resulting from the assignment of significance to otherwise non-significant phonic or graphic material. The alternative thesis would be that words originally attained meanings from palpable or natural motivations, e.g., by registering or imitating the perceptible sounds or (in writing) shapes of the things to be signified, as in onomatopoeia and hieroglyphics.

1.3.17 Simple versus complex. A theory might work chiefly with ostensibly simple cases, such as the meaning of proper names, in the hope that others can be treated as straightforward “complications”. Or, a -theory may attack complex cases, such as the meaning of .the “self”,. on the grounds that these are qualitatively different from simple ones, not just quantitatively.

1.3.18 Legitimising versus critical. A theory may seek to legitimise a particular set of meanings, e.g. those of scientific terms, at the expense of others, e.g. those of emotional experience. Or, a theory may seek to enable critiques of any set of meanings as just one domain among many alternatives.

1.4 If we plotted these parameters on a graph and entered a large number of past semantic approaches, we might notice some conspicuous clusterings. For example, the vast majority of approaches, even those based on the “empiricist” school of philosophy, have been derived chiefly by speculation. The notion of empiricism as program far observing and describing the activities of representative groups of people is comparatively modern and has had limited impact on the field of semantics, doubtless due to problems in gathering and interpreting data in ostensibly “rigorous” ways.

1.5 The search for general theories of meaning has long been predominant and was reaffirmed by modern “ general linguistics”. Wittgenstein (1958: 18) surmised that “ the craving for generality” had promoted a “contemptuous attitude toward the particular case”. However, a general theory cannot be adequately supported by instances of general meanings only, whish are prone to seem vague and mutable in content and reference (e.g. the meaning of “thing”). On the other hand, specific meanings (e.g. for “pancreas”), though easier to circumscribe, are less commonly invoked. “Proper” names given to a single person or place are the most specific in reference but the least specific in ideational content and thus quite unlike “common” words.

1.6 A similar quandary obtains for abstract meanings (e.g. for “intellect”), which should lend support to the abstractions required to construct a theory, but which make less convincing demonstrations than concrete meanings do (e.g. for “chair”, 3.27). One might argue that all meanings were originally concrete, but this argument would not explain abstraction itself and would be hard to document (cf. Beaugrande 1988a).

1.7 Post-Saussurian linguistics has also fostered distinct preferences for deterministic, static, and synchronic accounts of meaning. If meanings can be shown to have just these criteria, they can be analysed and described for once and for all, thus guaranteeing steady and cumulative progress in semantics and simplifying the semanticists’ job. But if meanings do not conform to these criteria, the job is made much harder by the semanticists’ own efforts to impose them.

1.8 Denotative and componential models entail the comforting “building block” postulate that meanings can in principle be dismantled into more basic elements. At the “core” of the meaning would be the essential non-decomposable elements establishing the definitive value and identity of the meaning against all other meanings. Putting this postulate into practice, however, can be another precarious job. Finding and labelling the core would be the hardest: if one holds all meanings to be decomposable in principle, non-decomposable core meanings should be anomalous (cf. 3.6ff, 3.23).

1.9 The concept of arbitrariness had been introduced to disavow causal or motivational relations between word and meaning, or between “signifier” and “signified” (1.3.16). Any potential discontinuity implied by the lack of such relations was compensated by Saussure (1966 [original 1916]: 113, 14) when he cited “the arbitrary nature of the sign as an “explanation” of how “the social fact can create a system”, and compared “the social side of speech to .a contract signed by the members of a community.” He also argued that the “arbitrary” quality is offset by the “differential” one, by the “non-coincidence” of any “segment of language with the rest” (1966: 118) (cf. 2.4). The absence of “necessary relations between sound and meaning was further adduced as a reason why “in linguistics, to explain a word is to relate it to other words” (1966: 189) (cf. 3.7).

1.10 Yet Saussure suspected “arbitrariness” of being “an irrational principle” which could “lead to the worst sort of complication” (1966: 133). He was thus “convinced”: “everything that relates to language as a system” contributes to “the limiting of arbitrariness, and we must study “language” from just this standpoint (ibid.). He seemed to have pinned his main hopes on grammar, e.g. when he contrasted “grammatical languages” having .a “minimum of arbitrariness” with “lexicological” ones having a “minimum of organization” (ibid.). As for the lexicon, Saussure offered a wishful comparison of language to “a dictionary of which identical copies have been distributed to each individual” (1966: 19). If this comparison held, every speaker -- and hence every semanticist -- would indeed attain the same meanings, and these could afford to remain at any degree of arbitrariness (could be arbitrarily arbitrary, so to speak).

1.11 In my impression, the conception of “arbitrariness” as the essence of the relation between language and meaning has been a continuing source of uneasiness. By uncovering the systematic and stabilizing aspects of meaning, semantics could replace some degrees of arbitrariness with some degrees of motivation and functionality. So far, however, these aspects have been chiefly sought in the meanings themselves, rather than in the uses of meaning, although the uses are most plainly not arbitrary (cf. 3.22).

1.12 Though every science claims to be objective, the claim is easiest to defend when the objects of inquiry are also “objects in the world”, e.g., the rocks of geology or the animals of zoology. Meanings can be handled as objects only through special tactics, such as selecting meanings whose standard “referents” are “real objects” (e.g. “chairs”, 3.27); or, more subtly, by displaying meanings in spatial graphics with dimensions and subdivisions labelled “components”, “semantic features”, “sememes” and so on (cf. 1.17; 3.5). The subjective qualities of meaning, which after .all cannot exist apart from the participation of a subject, have generally received 1ess attention, especially when the “scientific” quality of the investigation was emphasized (e.g. by Bloomfield 1933).

1.13 Perplexities like those sketched above might indicate why “modern “semantics” has been subjected to complex pressures and has responded with characteristic arguments and tactics. No other field has quite the same concerns and problems, but commonplace attitudes about science, generality, objectivity, and so on, were bound to influence the agenda. We can therefore appreciate why the projects proposed for semantics in 1.1 have been hesitantly or unevenly treated. Such issues as the nature of meaning and its relations to knowledge and language seemed too intractable to be attacked head-on with currently accredited methods. Instead, semantics often preferred the more circuitous tactic of focusing on particular aspects of meaning which might eventually add up to an explanation or theory in some still-to-be-defined sense. However, such a process of addition presupposes some inner uniformity and coherence among individual proposals, whereas, for reasons I shall examine later, diversity and confrontation have been predominant (cf. 2.1f, 2.7, 3.10, 3.3, 3.14).

1.14 The importance attributed to the project of designing representations of meaning should be seen in this context. Such a representation might lead toward uniformity if it provided a consistent and complete repertory of descriptive terms and of constitutive rules for applying, relating, or combining them. Such demands are of course unrealistic if we expect to find such a repertory “in” the meanings themselves, which are, on the contrary, often found to be inconsistent, incomplete, and resistant to rules. Instead, sudden demands can only be met, if at all, in the uses of the re-presentation, i.e., in the activities of coding or recoding meanings into it or out of it.

1.15 At this point, potential relations between “meaning”, “language”, and “knowledge” assume a new urgency. If we assert that meaning is dependent on knowledge but not on language, we should design a language-free representation for knowledge. Or, if we hold meaning to be dependent on language but not on knowledge, we can design a second-order language (a meta-language) for representing meaning, e.g. by standardized paraphrasing or, if we assert that meaning is independent of both language and knowledge, our representation can be a purely formal -- i.e. both non-psychological and non-linguistic -- abstraction, such as predicate calculus or successor arithmetic.

1.16 However, any such dependence or independence is so hard to demonstrate conclusively that we might have to postpone semantics indefinitely if we waited for the results. Meanwhile, we may have to be content with impressionistic or anecdotal arguments. Meaning must be at least partially independent of language if meaning can stay reasonably constant during translation and if people can have ideas, concepts, or mental images in consciousness without necessarily having them coded in particular language expressions. Meaning must be at least partly independent of knowledge if every person’s knowledge is to same degree personal and idiosyncratic, whereas meanings are chiefly social and sharable. Yet these suppositions do not indicate the extent of independence and can hardly be proven because almost any proof would involve knowledge and language. We cannot seem to shed this involvement long enough to access a “pure” or .independent” meaning -- “to catch the meaning out of the air, yet to have it inviolably there, as my old friend, the American poet Richard Eberhart, put it (cf. Beaugrande 1985).

1.17 Most semantic representations presented so far have been volatile mixtures of ordinary language, specialized meta-language, and formal symbols}s, e.g. alpha-numerics, quantifiers, Greek letters, and arrays (e.g. matrices, rewrite-rule tables, truth-value tables) (cf. 3.8). Such representations would work best if meanings are static and deterministic enough to be recoded without significant alteration. If meanings are essentially non-deterministic, dynamic, and participant-centered, however, such recodings become a crucial part of the constitution of the meanings and must be accounted .for (cf. Beaugrande 1988a).

1.18 The question can then be raised whether and how far the recoding and the analysis and synthesis it requires are similar to the typical human processing of meaning (2.3f). Like other speakers of the language, the semanticists presumably share the general knowledge of the language, the society, and the world. But the goals of semantic analysis differ conspicuously from the goals of everyday communication, e.g., in a heightened readiness to perceive samenesses and differences (cf. 3.3f, 3.7). Moreover, each meta-language or formalism exerts pressures on the recoding to foreground certain aspects and to background others (2.4). For example, if quantifiers are introduced, the recoder will want to determine whether all instances of the referent are entailed (universal quantification) or if only one such exists (existential quantification). Suppose we want to understand the meaning of a text-segment like this:

 (1) JIMMY: They all want to escape from the pain of being alive. And most of all from love. (Look Back in Anger, p. 93, emphasis added)

They previous discourse contains no candidate referent for “they”; only one plural noun appears in the dialogue for the whole scene, considerably earlier and obviously not relevant here (“dark plots”). Since the main topic of the previous speech has been two women getting ready to leave, “they” presumably refers to them, but what of the “all”? Logically, it could be a “universal” quantifier for a statement about women, but precisely this universality is easy to interpret as a typical hyperbole of Jimmy Porter. Thus, to recode with a quantifier would be to sanction the hyperbole and cancel the tension between the presumed content of the statement and the warrant for it within the ongoing discourse interaction.

2. Meaning as parametric adjustment

2.1 The projects and parameters I sketched in section 1, though certainly not exhaustive, are typical for the context of activities which count as “doing semantics” and for the implicit decisions entailed in doing so in one way rather than another. The major reason why semantics has been lacking in unity and progress, I would submit, has been the attitude that not just one’s theory or model, but meaning itself must be definitively plotted at one point on each parameter, i.e., that our task is to state, for once and for all, just how general or specific, abstract or concrete, etc., meaning is by its very nature. We have not conducted adequate explorations of how meaning is continually adjusted in terms of generality, abstractness, and so on, depending on the circumstances of use. In that view, the activities of being meaningful always involve some adjustment of the parameters, some provisional but usually appropriate balancing of complementarities of which neither side is denied in principle.

2.2 In this account, the main difference between “being meaningful” in ordinary communication and “doing semantics” of the usual type wound be the ambition of semanticists to make definitive and theoretically validated adjustments. This distinction necessarily creates problems when these adjustments are projected onto the domain of meaning itself, thereby truncating the parameters and leaving the theory powerless to deal with aspects of meaning that reflect different adjustments. The theory becomes highly vulnerable to the presentation of non-conforming counter-examples. A predictable result could be the diffuse and unstable character of semantics and the confrontational quality of many semantic discussions (cf. 1.13; 2.7; 3.10, 3.13). Semantic theories claiming definitive adjustments must either come into conflict or else belie their own claims by practising mutable adjustments in order to accommodate counter-examples. This double-tracking may remain inconspicuous because semanticists, like everyone else, are so skilled in adjusting meanings to context that they are hardly aware of their own involvements and interventions (1.16).

2.3 It is therefore unlikely that semantics can be unified by the triumph of a single theory whose implicit or explicit parametric adjustments are proclaimed to be just the ones essential to meanings at large. Instead, major progress in semantics will require a widespread commitment to integrative models in which alternative positions or theories are coordinated and consolidated in terms of how their projects adjust the parameters and which aspects of meaning can thus be accommodated (3.13). Just as meanings are functional because they can be more or less determinate, more or less abstract or concrete, etc., so too can semantic theories be adjusted to complement each other rather than to compete with them or to refute them. The pursuits of semantics would then be more creditable models for the general social activities of being meaningful (cf. 1.18).

2.4 The “social fact” invoked by Saussurians could be interpreted not as the result of “each individual having an identical copy of the dictionary”, nor as the counterpart of “arbitrariness” and “difference”. or “non-coincidence” (1.9f), but as the consensus among different discourse participants making similar adjustments of meaning to fit relevant contexts. The issues of how meaning is acquired and used finally assume priority over the search for representations, because one’s acquisition and use of meanings enable and control one’s strategies for making adjustments. Unless we attain a clearer notion of those strategies, we will not have adequate means either to decide what we want to represent nor to determine what sorts of further adjustments we are making when we code a meaning into or out of a given representation (1.18).

2.5 The parameters listed in 1.3 are of course not exhaustive for semantic theories, but they are far less so for the meanings themselves. For example, meanings may be also positive or negative, emotionally intense or subdued, salient or non-salient (i.e. appealing to sensory impressions), and so on, whereas such parameters seem scarcely applicable to semantic theories. To be sure, we may be accustomed to rather narrow and idealized notions of “theory” that marginalize and conceal the values and feelings of the theorists. We might profit by investigating the emotional appeal or the sensory impact of theories, along with other parameters that prove to be relevant for meaning itself. But I doubt if the theorists under scrutiny would be grateful.

2.6 The parameters in 1.3 may help to identify some major adjustments that theories regularly perform on meaning. The last two can serve here for illustrations. Each theory is compelled to adjust the relative complexity of the domain (1.3.17; cf. Beaugrande 1987a, 1991a, 1991b). In the traditional -building block notion of complexity, a meaning would be complex if it had many parts, as maintained by “componential” approaches. But this notion itself is simple in implying that a meaning can be decomposed and can thus be treated as the sum of its parts, variously called “features”, “sememes”, and so forth (cf. 1.3.14, LB, 1.12). Inconclusive debates, however, suggest that decomposition is an unstable adjustment that different semanticists perform in diverse ways (cf. Bolinger 1965; Beaugrande 1984b, 1988a).

2.7 “Ordinary discourse” has a reputation for being simple, but detailed investigation, e.g. in “ethnomethodology”, indicates quite the contrary. The complexity of ordinary discourse is masked by the familiarity of its contexts and settings and by the detailed world-knowledge participants bring to bear when assigning meanings. We might hope that if we “abstract away” from contexts and settings we can simplify our description; but the complexity returns in the activities we must then perform to reconstruct the meanings of the “abstracted” materials without the contributions of contexts and settings to the control of complexity. The endless disputes about “semantic features” and the like are a striking instance.

2.8 Another strategic parameter would be the adjustment of relative authority. In theory, we can describe a meaning whether or not we believe it to be the content of an authoritative statement; yet we would like our description itself to be authoritative. Many past trends in semantics, notably those influenced by positivism, have understandably attempted to borrow authority for the description by admitting ostensibly authorized meanings for statements about physically verifiable reality (cf. 1.3.5; 1.12; 2.11). This loan becomes burdensome when it entrains the semanticist in a drastic epistemological overextension of a handful of special cases and a corresponding disinterest toward a host of other cases (cf. 3.10).

2.9 The authority claimed for a project may extend in two directions (cf. 1.3.1B). In one direction, legitimising projects approach semantics as a vehicle for authorizing certain interpretive practices and contents. One eminent instance has been theological postulates about “divine wisdom” as the origin and final arbiter of thought, meaning, and language. This legitimation offered the twofold advantage of shielding the ultimate source of meaning behind divine authority and of flattering humans with the capacity to mimic the divine. A counter-instance has been the search for the origins of language in biological evolution, e.g. as a means for .survival and as a reflex of bodily functions and movements, such as hand signals and facial expressions. This legitimation dispenses with the divine but flatters humans with the idea that their communicative faculties represent the culmination of long evolutional processes and the triumph of our intelligence over nature.

2.10 Although the decline of theologies has often been linked to the rise of “modern science”, the ambitions for legitimation have remained largely in place. As the authority of divine systems responsible for universal order and sense was increasingly called into question, new channels and procedures were urgently needed. Science undertook to provide them, but at the same time provided new practices for undermining legitimacy, e.g., experimental tests. This duality has entrained science in its own double-trackings, such as claiming full authority for the current theory even though the history of the science consists largely of refuted theories, or asserting all knowledge to be a product of concrete experience yet purporting to hold transcendent knowledge about all possible experiences within the world or the universe.

2.11 The meaning of scientific discourse is thus also open to peculiar pressures. The strongest pressure is to extend the claim of “objective” and “disinterested” authority over to the discourse of inquiry and description as well, e.g.:

(2) One of the accepted properties of science in its rigorous forms is that the language of discourse is objective in contradistinction to indexical. Scientific statements purport to be context-free (and) to have validity independent of the observer, independent of historical time, and applicable to all situations. (Attewell 1974: 185).

The “indexical” covers all references to the current situation of the discourse itself (cf. Bar-Hillel 1954). And yet that very situation might be viewed as the most realistic grounds for the discourse and the most reliable orientation for telling what things mean. So the scientific theory guiding the inquiry and description must act as a filter for removing the irrelevant realness and retaining only the relevant residue to constitute the meaning of the discourse. But it’s hard to see how such a filtering theory could “a pure observation language” from whose discourse we would have “eliminated” “all non-logical and non-perceptual terms” (cf. Kuhn 1970: 126f), thereby leaving only the terms whose meaning is derived either by derivation from the known value of others (i.e. logically) or from “some publicly accessible manifestation (i.e. perceptually). On the contrary, the removal of the “indexical” aspect would filter out most opportunities for unmediated perceptual derivation of terms and meanings.

2.12 The converse direction would be critical projects which alternate viewpoints in order to foreground the ways in which meanings are relative and sensitive to each viewpoint. Here, meanings are pursued to the stages where divergences or contradictions become steadily more obtrusive and the decisions to favour one version over another can be recognized to stand in the service of personal, ideological, or institutional priorities. No configuration of meanings can be neutral or self-evident; but habituated ones typically seem to be so because they so often function in the unstated background for the meanings a person has in focus.

2.13 Critical methods, however, are easily co-opted by an interest in legitimising a particular outlook by exempting it from one’s criticisms aimed it against its competitors. Also, originally critical methods can be turned into new orthodoxies, e.g., when Freudian and Marxist critiques began by undercutting conventional (“proper”, “obvious”, etc.) meanings generated in the service of social enforcement and ended up by asserting the untranscendable status of psychoanalytic and socio-economic meanings, respectively (cf. Beaugrande 1988b).

2.14 The notion of meaning as “parametric adjustment” would suggest that “critical” approaches should be well-suited for reflecting upon the activities of being meaningful, including those of “doing semantics.” However, we can also appreciate that critical approaches will be hard to develop and even harder to maintain. A “critical” method can remain so only as long as it is continually applied to itself and to the meanings it generates and describes, and perpetually and deliberately readjusts its own parameters to create fresh differentials (3.11). This desideratum may prove enormously demanding.

2.15 In addition, the requirements of a “critical meta-language” cannot be well understood as long as meta-languages are routinely constructed for purposes of legitimation (2.11). If we undertake to construct a complete new meta-language out of whole cloth, we would have to prefabricate the whole system of meanings for it, and how we could do so without using ordinary language is difficult to imagine. Endless regress impends, as in this proposal:

 (3) As scientists, we seek to produce a literal description of our subject matter. In order to describe, we construct (or adapt to our use) a language. While to begin with our language may be crude, one rule must be constantly attended to: nothing we take as subject can appear as part of our descriptive apparatus unless it itself has been described. (Sacks 1963: 2)

Sacks’ “rule” leads into a circle by suggesting we must have described our terms before they can be used in a description, although we would not yet have a language for doing the describing. At best, provisional descriptions in ordinary language could be critically recoded piece by piece into the meta-language until we have enough “critical mass” to enlist the meta-language in the production of more controlled descriptions and to deploy one part of the meta-language for describing another part. Even these compromises have seldom been pursued very far in practice, because scientists are eager to get beyond the meta-language itself into the domain of inquiry. In semantics at least, this eagerness has frequently shielded the meta-language against critical reflection.

2.16 If “meaning” as such were a fully independent and self-sufficient system, then every instance of “meaningful language” would already be meta-language superposed on that system. But since meaning is extensively integrated with language, we still want a meta-language “above” language, and we would prefer it to have a different order of integration with meanings -- supporting the “objective”, “valid”, “context-free”, and “literal” qualities invoked by Attewell and Sacks. Yet these qualities, being in essence legitimising ones, could undermine the potential of the meta-language for critiques of its own recodings and descriptions.

2.17 As far as I know, at any rate, semantics has never provided a complete meta-language which is at once above ordinary language and yet related to it in an objective, valid, context-free, or literal way. We might do better to pursue to the prospects for developing a meta-language which is not intended to have these idealized qualities, but rather to facilitate critical appropriations of meanings, including its own.

3. A challenge for text semantics

3.1 At this stage in the deliberations, we can turn to the problems and prospects of text semantics. The issues aired in the first two sections should indicate why we might have reservations about a “text semantics” derived by some extension of conventional “formal” or “linguistic” semantics. If we expropriate a “semantic theory” which proposes definitive adjustments and legitimises particular practises of sense-making, our representations for “text meaning” will be partial and inadequate. Either we rarefy and reduce the text meaning to conform with the predetermined “generative powers” of a standard grammar and lexicon; or we allow for rich text meaning by complicating and adjusting the grammar and lexicon to accommodate the text at hand, e.g., with massive additions of “rules” and “features”. Either way, the theoretical apparatus will play more a legitimising than a critical role, especially if, following the tendencies of logic and formal grammar, we assume that the meanings to be recoded and represented should be rendered fully determinate during the process, e.g., in terms of referents, truth values, syntactic structurations, and so on. Rules and features are intended to be descriptive, but, to the degree they privilege certain adjustments over others, take on a prescriptive function as well.

3.2 Conventional semantics (and linguistics as well) has cultivated a meta-language of non-involvement, wherein it has been customary to say that a certain word, phrase, or sentence “means” particular things in and by itself, as if under its own power (cf. 1.3.12). In a critical analysis, this meta-language might be examined partly as a. shorthand for jumping over our own interventions to get to the meaning and partly as a “loan” of stability we could in principle “repay” by demonstrating that when typical speakers of the language utter the word, phrase, or sentence, they normally mean “the same things” as we construct. However, the readiness of semanticists to carry out such demonstrations has not been conspicuous so far; quite apart from the legwork needed to ‘set up the tests, one would soon collide with the recursive problem of deciding what the data “mean” and what evidence should be accepted as an indication of the “same meaning” rather than a “different” one (cf. 1.4, 1.18; 3.7). The “loan” therefore keeps growing and remains unrepaid (cf. Beaugrande 1991a).

3.3 The notion of “sameness of meaning” is itself by no means straightforward. If meaning is readily adjustable, then an exact match between two instances of the meaning, even for the same words uttered by the same speaker, might be more the exception than the norm. Besides, memory research indicates that it is nearly impossible to distinguish between what is actually remembered and what is reconstructed” (Loftus 1980: 125). “New information to which a person is exposed seems to replace irrevocably the original information in a person’s brain”; “whenever a memory for an event is called to consciousness, the potential appears to be there for substitution or alteration” (1980: 49). Thus, exact reproduction of meaning would hardly be feasible, and adjustments would tend to be continually incorporated.

3.4 Problems also beset the concept of “difference”, which Saussure placed at the centre of linguistic theory and proffered as a counterforce to arbitrariness ( cf. 1.9). Indeed, he vowed that “in language there are only differences without positive terms”; and that the “function of the linguistic institution” is to “maintain the parallelism between two classes of differences, those among “signifiers” and those among “signifieds” (1966: 120f). But it would surely prove disputatious to declare at what degree the adjustments performed on a given meaning from context to context justify a judgment of “difference” in this sense. It would be more plausible to assume that each reactivation of a meaning involves an adjustment between sameness and difference respecting past activations, and that the current meaning is crucially determined by these adjustments rather than by the absolute differences Saussure envisioned at the level of the abstract language system (3.7). We should envision sliding ratios of sameness and difference and relative scales of consistency; realistic measures will have to be empirically determined and will, I suspect, show considerable variability (3.28). In addition, semantic analysis probably enhances sameness and difference for purposes of classification, determination, and so on (1.18).

3.5 As a chief principle, text semantics might find productive approaches to long-standing problems by viewing “text meaning” as a systemic configuration of parametric adjustments along semantic dimensions. These adjustments are “systemic” to the degree that they are rendered mutually relevant during the utilization of the text. The openness and flexibility of word meanings as “lexical entries” is the strategic prerequisite and correlate of the capacity for adjusting the dimensions in appropriate contexts (3.27). The lexicon therefore cannot be expected to contain either definitive or exhaustive specifications any more than the set of “all possible contexts” can be anticipated. Parametric interactions can always modify values in unusual ways. For sample (1) in 1.18, the meanings for “being alive and “love”, which are typically positive, can be adjusted toward negative values to come closer to the meanings of “escape” and “pain”. One might indeed conjecture that a central intention of the whole drama text Look Back in Anger is to work its characters and the audience through a gamut of ambivalent emotions with dizzily fluctuating values and interactions, such as Jimmy’s systemic associations between affection and disgust:

(4.1) Give her her finger back and don’t be so sickening (p. 13)

(4.2) You both look pretty silly slobbering over each other [...] You’re making my stomach heave (p. 31f)

(4.3) my wife [...] sweet and sticky on the outside, sink your teeth into it, inside, all white, messy, and disgusting (p. 49)

(4.4) we must have been married I suppose. I think I remember being sick in the vestry. (p. 55)

(4.5) “I shall always have a deep, loving need for you. --- Alison”.Oh, how could she be so bloody wet! Deep loving need! That makes me puke! (p. 72)

The conversion of desire to loathing was described by Freud as a typical one, but hardly seems a proper part of a lexical definition of “love” -- an item whose “true” or “core meaning” is notoriously elusive and open to adjustment.

3.6 We need hardly be surprised at the difficulties of telling which the “true” or “core meanings” are and which meanings are produced by adjustment if, as I assert, discourse is not predominantly a modality for conveying “true meanings”, but one for adjusting and correlating meanings. Certainly, it would be unreasonable to demand that text semantics must declare the “real status” of “true meanings” versus “adjustment meanings” before research can proceed. Extensive empirical explorations must be performed before we can determine whether and in what sense these conceptions may be “meaningful” as contrasts or complementarities. The belief in “true meanings”, though well-anchored in traditional speculations and commonsense reasoning, as well as in theology and science (2.9f), rests upon provided scant evidence from semantic research, which has routinely treated it as a postulate a priori rather than as a hypothesis to be challenged. We might equally well contemplate whether all meanings were not originally generated by adjustments, so that the “true” or “core meaning” of an expression is a pattern of habituated adjustments that have come to be executed as a package of operations when the meaning is activated in discourse. Still, such operations too must be in principle sensitive to the others performed for the same text, and the interaction of operation packages could generate the frameworks known as “context”, “coherence”, “continuity”, “topic”, “superstructure”, and so on (cf. Beaugrande 1980, 1984; van Dijk 1980; van Dijk & Kintsch 1983). Empirical probes of the operations can provide indications of the ratios between habituated adjustments and contextually adapted ones. The ratios probably vary according to the extent that the meaning and content do or do not match one’s prior knowledge and expectations (3.17).

3.7 If all meanings are ultimately the products of adjustments, we might attribute a new sense to Saussure’s (1966: 189) thesis (a correlary of “arbitrariness”) that “in linguistics, to explain a word is to relate it to other words” (1.9). Meaning would be a relational entity not merely because, as Saussure argued, its value arises in a system of differential elements, but because it is constituted by adjustments along parameters between values that are both similar and different, and by the interaction of such adjustments during production and comprehension of discourse (3.4). If so, the search for the ultimate, unchanging “core meaning” can lead only into further patterns of adjustment, including the adjustments made during the search itself. The choice of a final resting place would be “arbitrary” in a sense quite compatible with Saussure’s, but uncomfortable for semantic analysts.

3.8 Semanticists who construct a table or matrix of “semantic features” and the like are encouraged by the notation to select and label ostensible components and to mark these as either present or absent (with a plus or minus sign). The static, often binary construction of the formalism offers no occasion to represent a meaning as a point on a scale of more or less, positive or negative, abstract or concrete, and so forth. Thus, the semanticists are impelled to disregard the potential for adjustment, along with its correlates such as nuances, ambivalences, multivalences, contextual influences, and so on. The semanticists presumably intend their table or matrix to represent only “the abstract lexical meaning” of the word and hence the unadjusted core. But as I remarked (3.6), the thesis that all or even most lexical entries have such a core is far from certain; and in any case the core would very seldom be actualised in “pure” form in ordinary text or discourse and would thus scarcely be dominant enough to Justify its centrality in semantic theories and representations.

3.9 Networks might be an alternative formalism, one frequently recommended for instance by Halliday. He visualizes semantic networks” as a modality whereby “systemic theory” “interprets meaning as choice” and “a language” or “semiotic system” “as networks of interlocking options”; “whatever is chosen in one system” leads to “a set of choices in another. as we move from “the most general features step by step to the “specific” (1985: xiv). Halliday has made broad claims for such networks, e.g. that they can provide an “account of how social meanings are expressed in language”, and “a description of each meaning selection and an account of its relationship to all the’ others” (1973: 65, 88), The choices would presumably appear as branchings between options, only one of which can be selected to the exclusion of the others, e.g., when Halliday says: “a system is a set of features, one of which must be selected if the entry condition is satisfied” (1967: 37; 1973: 55), So adjustments could hardly be represented, because they imply intermediary routes between alternative branches and conditions.

3.10 Contemporary “philosophy of science” has acknowledged that the supposedly objective and disinterested content extracted directly from experience, as foreseen by classical science, must be adjusted by the interaction of perception with identification, expectation, interpretation, etc., e.g.:

 (5) The content of a concept is determined [...] by the way in which it is related to perception. Yet how can this way be discovered without circularity? Perceptions must be identified, and the identifying mechanisms will contain some of the very same elements which govern the use of the concept to be investigated. We never penetrate this concept completely, for we always use part of it in the attempt to find its constituents. (Feyerabend 1978: 76).

Adjustability is acknowledged for scientific discourse as well, although the operational adequacy of understanding is maintained, e.g.:

 (6) The notion of an expression that has a fully specified meaning or of an expression that is fully understood is obscure; besides, even for terms that are generally regarded as quite well understood there are open questions concerning their proper use [...] Theoretical concepts [...] are “open-ended”; but that, evidently, is no bar to their being adequately understood for the purposes of science. (Hempel 1970: 163).

The most severe problems and confrontations arise when either classical or non-classical positions are pushed to extremes. The prescriptive thrust of the classical position is anchored in the supposition that the proper uttering of a potentially “universal statement is precisely the one that realizes the intention of universal applicability. When such an uttering is then over-generalized as proof of universality itself, counter-arguments are easy to construct, e.g.:

(7) Can there be statements which are applicable to all situations? Bar-Hillel’s (1954: 359) classic example of an objective expression in his sense of this term is “Ice floats on water.” Outside of the various physical constraints on the truth conditions of this statement and the infinity of tacit etceteras which would need to be added in order to guarantee its universality [...] we can imagine contexts in which the use of this utterance is quite different from that envisaged by Bar-Hillel for himself and for almost every grown-up normal English speaking person.. An example might be an ironical use [for] a person known for their frigidity bathing in a pool [and] floating (McHoul 1982: 89f)

But the other extreme is equally vulnerable, namely when an uttering in the service of a non-universalising intention is overgeneralized, e.g.:

 (8) That there is no utterance which cannot be heard, read, or otherwise used as ironical, metaphorical, non-serious, etc., on particular and appropriate occasions vitiates the possibility of singularly objective expressions [...] “applicable to all situations”. (McHoul 1982: 90).

McHoul’s disclaimer of universal applicability is itself advanced with universalising diction: (“there is no utterance”, “vitiates the possibility”). The option of “ironical, metaphorical, non-serious cases” can “vitiate the possibility of singularly objective expressions” only if we retain the questionable provision that we can generalize from every case of meaning “on a particular and appropriate occasion” over to the set of possible meanings. McHoul seems to replace Bar-Hillel’s absolute of seriousness with his own absolute of non-seriousness. Surely seriousness is one important parameter for a wide range of adjustments.

3.11 Such disputes at the very foundations of epistemology and discourse again indicate that text semantics should not merely explore the practices of mean