Linguistica Pragiensa 34/1-2, 1992,2-26 and 55-86.

 

THE HERITAGE OF FUNCTIONAL SENTENCE PERSPECTIVE

FROM THE STANDPOINT OF TEXT LINGUISTICS{1}

 

Robert de Beaugrande

 

1. Formalism, structuralism, functionalism

 1.1 From the standpoint of text linguistics today, the central theses of the Czechoslovakian “functional” approach, some of them articulated over sixty years ago, sound strikingly modern. Whereas linguistic schools like those in Geneva, Paris, and Copenhagen exalted the “structural principle”, formulated particularly by Saussure, that a “language” is an abstract system of mutually related and determining elements, Czechoslovakian schools like those in Prague, Brno, and Bratislava promulgated the “functional principle” that a “language” is a system of elements to be deployed in communicative functions. And whereas Saussurian structuralism embraced the thesis that “the true and unique object of linguistics is language [langue] studied in and for itself”, a “fundamental thesis of functional structuralism” was that “language [Iangue] is a system of means appropriate to a goal”.{2}

1.2 According to Vilém Mathesius, the chief founder of the Czechoslovakian functional approach, “the traditional method of linguistic research may be called 'formal” in the sense that the form as the thing known has constantly been made the starting point of the investigation, whereas the meaning or function of the form has been regarded as that which should be found”.{3} This tradition doubtless reflects the impression that the forms of language, once they are transcribed, constitute much more tangible manifestations than do meanings and functions. However, to the degree that language is a medium expressly designed for organizing and correlating forms, meanings, and functions, the limitation of a description to the transcribed and abstracted forms is more problematic than has been widely believed. In effect, the investigator utilizes meanings and functions during the process of discovering or arranging the forms, but does not disclose these uses in the final description. In consequence, the uses of meaning and function as an integral part of both language communication and language description have not been subjected to adequate examination and control. This problem becomes acute whenever two or more formal descriptions of the same system or phenomenon cannot be reconciled; we would need an explicit account of how they were attained through diverging appropriations via meaning and function.

1.3 Mathesius (Ioc. cit.) programmatically advocated a “modern linguistics” which “takes the meaning or function as the starting point and tries to find out by what means it is expressed”. Whereas “older linguistics” “proceeded” “from form to function”, such a “modern linguistics” would “proceed from function to form”. In actual practice, though, neither direction of “procedure” could be exclusively maintained. Whereas any formal method must consider function without acknowledging it, any functional procedure must appropriate forms as a substrate for implementing and identifying functions. Thus, the two procedures depicted by Mathesius should be two complementary agendas or priorities.

1.4 But the history of modern linguistics reveals a periodic and unproductive antagonism between formal and functional agendas. The problem remains: how and how far can forms be separated from the functions they carry in actual occurrences without thereby sacrificing the key qualities that constitute language? Most versions of formalism involve quite arbitrary assumptions on this point or evade the problem with ambiguous notations that are explicitly formal but implicitly functional (2.9, 19). Yet until the problem is squarely faced and resolved, we cannot precisely determine how formal constructs and the conclusions based on they correspond to whatever facts we can derive from the data of language functioning in real-life discourse. Many correspondences are possible in principle, and the selection of any one may significantly influence the results.

1.5 The two agendas project sharply diverging modes of analysis. Whereas a formal analysis is progressively constrictive and terminates at a set point when 'all forms are identified and connected, a functional analysis is progressively expansive and shades off in degrees of decreasing relevance (cf. 2.22, 30). The analyst as language user is heavily involved in both kinds of analysis, but formal methods render this involvement less obtrusive. Functional methods are more likely to point to the complex and context-sensitive dispositions and decisions the analyst must undertake in order to attain a given result. This effect may leadlinguists to decry functioned analysis as subjective, idiosyncratic, and so on. Yet in fact the contrary may hold: formal analysis obliges the analyst to rely on arbitrary, untested suppositions and decisions about the roles of particular forms. Formalism does not eliminate the “guesswork” charged against functionalism (3.67), but introduces its own mode of guesswork (1.2, 18; 2.20, 31).

1.6 The central thread in the history of linguistics so far has been a continual, circumspect manoeuvring along the borders of forma Formalism proper tries to keep this manoeuvring on the formal side as far as possible and to cross over into function only for brief and proximate incursions. We see the influence of this ambition in the general assumption that function and meaning are genuinely secured when we can find distinctive formal correlates and consequences. An English grammarian for instance expects a classification of “sentence types” to include those that are formally differentiated in English, such as “question” and “exclamation”, whereas a linguist dealing with other languages, such as K.L. Pike, might postulate types like “disappointment”, which was inspired by its formal differentiation in Menomini.{4}

1.7 Due to this concern for formal anchoring, the most influential experience for structuralist linguistics was its engagement with languages that correlate form with function In ways quite unknown in familiar ones like English. One pioneer, Edward Sapir,{5} opined that “a sympathetic grasp of the expression of the various classes of concepts in alien types of speech” is encouraged “when one has learned to feel what is fortuitous or illogical or unbalanced in the structure of one's own language”. “It is often precisely the familiar that a wider perspective reveals as the curiously exceptional”. The highly diverse correlations between form and function in, say, Amerindian and Polynesian languages reveal the familiar descriptive terms for English to be far from exact or transparent. In Yana of Northern California, Sapir was perplexed to report, “the noun and the verb” “hold in common” some “features” that “draw them nearer to each other than we feel to be possible”; and the language has, “strictly speaking, no other parts of speech”, “the adjective”, “the numeral, the interrogative pronoun”, and “certain conjunctions and adverbs” all being “verbs” (op. cit., 199). If the correspondence between form and function had to be the same in all languages as in English, then Yana would suffer from such a severe functional underdifferentiation that communication would scarcely be feasible. The so-called “Sapir-Whorf hypothesis” that each language leads its users to see the world differently might be restated: for various reasons -- only some of which are historically, culturally, or psychologically proximate or discoverable -- each language correlates the organization of forms with the organization of functions in its own particular ways. To compare languages, we need a terminology and description based neither on forms alone nor on functions alone, but in form-function correlations.

1.8 Functionalism too manoeuvres along the borders of form, albeit on the other side from formalism. Michael Halliday puts it this way: though “based on meaning”, “a functional grammar” is “an interpretation of linguistic forms”: “every distinction” -- “every set of options, or 'system'“ -- must “make some contribution to the form of the wording”.{6} He set forth the “principle” “that all categories employed must be clearly 'there' in the grammar of the language”, “not set up simply to label differences in meaning”.

1.9 In traditional grammar -- and much of the criticism linguistics has directed against it reflects this very factor -- the complementarity between form and function was routinely accepted at face value without much self-consciousness or dispute. Commonalities of function were seen as a handy resource for stabilizing a formal category. A category like “dative” or “ablative”, for instance, which occupied a certain position in each declension of Latin, was centred on the functional conception of “giving” or “carrying away”, respectively, just what the etymology of the terms suggests. These functional identities preserved the order of the declension despite the widespread formal convergence of dative and ablative in the plural. In contrast, formalist linguistics went further by suggesting that the functional import or meaning of such formal category was (at least originally) obligatory and exclusive, whence Jakobson's thesis of invariant meaning for each “case” in Indo-European declensions.{7}

1.10 Yet functional support had to be flexible so that historical changes in whole systems of forms would not render the system chaotic. In comparison to Old English, for example, Modern English has leveled most of its inflectional paradigms of declension and conjugation, and some dialects have carried this trend still further than has standard written English. In return, word order has become much less flexible, as particular arrangements have taken over the functional load previously carried by inflectional differentiation. The lack of formal distinctions hardly impedes communication, but obliges every formal linguistic description of English to incorporate functional information in some guise. A linguist's evergreen like sample (1) is easily described as “Article + Noun + Verb + Article + Noun” even though by themselves, man and ball could be verbs, and hit could be a noun; the formal description is supported by a plausible functional organization of “Agent + Action + Object of Action”. The familiar garden-path sentence (2), in contrast, can seem formally ambiguous on first reading because the functional organization is much less plausible, the old man being readily identified as an operative unit, whereas the activity of manning is far more uncommon than hitting and hardly suitable for old people anyway. Sammie (3) is less likely to be prematurely misparsed because tiring is a predictable action of the old, and an old tire makes a less reasonable agent for an action. Sample (4) would require a very peculiar context to indicate how the hit might be a group of agents and how they might go about manning a ball.

(1) The man hit the ball.

(2) The old man the boats.

(3) The old tire more easily than the young

(4) The hit man the ball

Formal linguistics is usually content to diagnose alternatives or ambiguities and to account for each reading as a correlate of differing but unambiguous forms, e.g., as an ambiguous “surface structure” for more than one unambiguous “deep structure” (cf. 1.23, 2.9f, 37; 3.22). This method again utilizes functional criteria to construct plausible formal correlations but describes them as formal criteria (cf. 1.2, 13).

1.11 All In all, modern linguistics has persistently underestimated the role of function and meaning in the appropriation of forms. Linguists have criticized traditional “parts of speech” schemes for relying informally on function and meaning, but their own formal schemes merely made the reliance implicit. The traditional definition of a “noun” as “the name of a person, place, or thing” certainly seems less objective and reliable than the formal definition of a “noun” as an item which can be preceded by the and followed by a plural ending. Yet “article” and “plural” too have not just formal aspects, but functional and semantic ones which mediate against applying them to certain nouns. Unless a special context is constructed, (5a) seems acceptable and (5b) and (5c) do not, even though mankind is undeniably a noun in all the samples:

(5a) Mankind is rapidly destroying the Earth.

(5b) The mankind is rapidly destroying the Earth.

(5e) *Mankinds/menkind are rapidly destroying the Earth.

The terms “definite” and “indefinite article” are even more conspicuously functional, indeed multifunctional (3.30) -- so that formal grammars usually prefer just to list them as a conveniently small and recognizable word-class labelled “ART”) rather than to offer a formal definition. Historically, one might formally classify the definite article among the deictics, and the indefinite article among the numerals. But in a contemporary (“synchronic”) description, we would have to admit that both form and function have changed since then; and the small repertory of forms corresponds today to a wide repertory of functions.

1.12 The agenda of modern linguistics has been dominated by the project of steadily increasing the formality of both theory and description and expecting that ail relevant functional factors would eventually be reconstructed as formal ones. The “distributional” linguistics advocated by Z.S. Harris proposed to consider only the arrangement of formal items in language samples.{8} A description enumerating which forms are regularly preceded or followed by certain other forms of course reflects meanings and functions as well. However, the formal description tends to underrate the fact that distribution (word order) itself is a meaningful and functional aspect of language (cf. 1.10; 2.18-22). If we say that man is a noun In Sample (1) by virtue of its systematic distribution, we are in effect saying that the form-function assignment is systematic enough to give rise to such a distribution in the first place.

1.13 Due to the dominant agenda, the immediate descendant of distributionalism was not functionalism but a still more elaborate formalism. Harris had realized that he could enhance the distributional regularity of a discourse sample by transforming its constituent sentences into parallel forms. If a sample discourse contained sequences like you will be satisfied and satisfied customers, the second could be transformed into customers are satisfied.{9} His star pupil Chomsky transposed this essentially heuristic tactic into a central principle for relating linguistic theory to linguistic data. A set of rules able to convert any type of allowed formal sequence into any other type would be a complete account or “grammar” of a language (2.2). Intriguingly, major support for this proposal was drawn from data in which forms and their distributions are unenlightening and underdifferentiated. Thus, the equivalent distributions of

(6) John is eager to please

(7) John is easy to please.

were held to conceal the grammatical (actually functional) aspect that John is doing the “pleasing” in (6) but not in (7). Since Chomsky also inherited Harris' aspiration for a “purely formal”{10} approach, he too failed to acknowledge that his recognition of the differences between such sentences was derived from intuitions about meaning and function. His “transformational grammar” and much of the linguistics it inspired was officially conceived as an enterprise for correlating not form with function or meaning, but form with form. The axiomatic basis of the theory was the “sentence”, not as a unit marked off in discourse on semantic and functional grounds but as a formal unit related by form-manipulating rules to every other such unit in the language. A full set of such rules which would “generate” all allowable e types and no others was projected to be the ideal tool for a” complete grammatical description (cf. 2.9). However, the constraints on what a real speaker might regard as a possible sentence proved formally intractable because they are prominently semantic and functional, and the ambition for a. purely formal account had to be modified. The revisions of Chomsky's “standard theory”, such as “case grammar” and “generative semantics”, were essentially agendas for importing semantic and functional criteria into a system of form-form correlations with the least possible sacrifice of formality (cf. 3.22, 25).

1.14 Many disputes have accordingly centred neither on the nature of language nor the status of data, but on the proper criteria and degree of formality a theory or description ought to have in order to count as “serious”, “rigorous”, “scientific”, and so on. The formalist stance has consistently been that only formal linguistics may establish these criteria, and that a theory may be superseded only by a more formal one; to move over to a less formal one would constitute a sacrifice of principle and a betrayal of science. The stance gives rise to endless controversy as long as we still lack a complete formal account of any language from which confirmed criteria could in fact be determined. Instead, we have only partial and fragmentary formal accounts resting upon considerable -- and as yet undetermined -- amounts of unacknowledged semantic and functional information.

1.15 Our problems are intensified by the prevailing interpretation of formality as a determinate phenomenon (cf. 3.28f). Because the forms of a language constitute the aspect of the language about which grammarians and linguists are most likely to agree, a formal description is expected to be deterministic it should, through formalization, eliminate all approximations, undecidables, and ambiguities, and should admit no equally valid alternative. However, language can be operational for communication only if determinacy is a fluctuating and distributive factor.{11} It belongs to the meaning of meaning and to the function of function that both meaning and function must be open and multiplex at the system level so that their correlation with forms can be determined in a wide range of ways at the discourse level.

1.16 The Czechoslovakian tradition going back to Mathesius presents an instructive contrast to the formalist agenda. If, as he counselled, we “take the meaning or function as the starting point and try to find out by what means it is expressed”, we have no a priori or definitive criteria for determining which mode and degree of formality and determinacy will be appropriate to the description. Since “function” is itself a correlation of factors, its nature is fundamentally both complex and non-deterministic. We may well appreciate why Mathesius observed that “the deeper insight we get into the organization of language, the more we are persuaded of its complexity and of the impossibility of arriving at” “clear cut statements”; or why Danes remarked that “the empirical study of languages shows clearly that any neatly patterned scheme of the structure of language does not sufficiently account for their complexities and semi-regularities”.{12}

1.17 It might help here to clarify a dualism within the term “complexity”.{13} In a positive sense, a phenomenon (notably a scientific theory or a work of art) is called “complex” when it succeeds in integrating multifarious and diverse aspects. In a negative sense, a phenomenon is called “complex” to suggest that its aspects resist integration. The chief business of science is to convert negative complexity into positive complexity. In its search for formal and deterministic description, linguistics has conventionally seen formal complexity as positive and functional complexity as negative. A chief tactic of description has thus been formal complication plus functional simplification. This tactic leads to imbalances and disturbances when the neglect of functional criteria allows the formal complexity of description to rise unchecked, e.g. in a massive complication and proliferation of formal “rules” and “features” (cf. 2.14).

1.18 The term “structure” also needs to be reassessed. If a “structure” is defined as “a relation between mutually relevant elements”, and a “function” is defined as “a role assigned to an element within a systemic configuration”, the two terms seem quite compatible. A given element might be said to have a certain “function” within a “structure” or a class of “structures”, while a “structure” could be designated as the product of functional assignment. However, the two terms have often been deployed in ways reflecting the theoretical and methodological disparities between formal structuralism (including generativism) and functionalism (sometimes called “functional structuralism”).{14} Typically, the “structure” was grasped as a bond between forms, and the “function” as an alignment between form and non-form. The “structure” was therefore more likely to be discovered and treated as a self-sufficient inter-formal relation within the language system, thereby downplaying the involvement of the linguist in the production of structures, or (in the aphorism of Andre Martinet, another prominent functionalist),{15} the process whereby “the structuralist is not one who discovers structures, but one who makes them”. “Structuralism” accordingly claimed to be more formal and rigorous than functionalism in theory, but in practice tended to be more incomplete and reductive.

1.19 The formal orientation naturally encouraged a restriction of the term “structure” to relations within a single subsystem or level, while “constituency” was said to obtain between elements that figure inside one unit on the next higher level (e.g. between “words” in a “phrase”) (2.29). The most exemplary “structure” was the “binary opposition” between two elements on one level. Here, the function of the elements was reduced to the absolute minimum: two elements of equal status (e.g. two “phonemes”) differing in a single crucial property (e.g. “voiced” versus “unvoiced”). The attraction of this concept encouraged radical theses, such as the absolute primacy of binarism and the strict separation of levels.{16) Yet to state that a form has the “function” of being an element in a formal structure, is hardly more than to reassert that it is a form.

1.20 For the reasons given above, we might prefer not just to attempt a functional reinterpretation of already-complicated formal linguistic theories and data, but also to inquire how functional and formal complexity can be manageably correlated in principle. If we work all the way through one side and then try to add on the other, the correlations tend to seem unwieldy and makeshift (e.g. as an occasion for another whole set of formal “rules” such as “transformations”, cf. 2.14). Only if we consider the correlations” at every step of discovery and description can we recognize their systemic interdependence.

1.21 In another contrast depicted by Mathesius,{17} “transferred into real life, the formal method coincides with the method of a hearer, who has to find the meaning of words or sentences he hears”; “the functional method” should consider “the point of view of the speaker or the writer, who has to find linguistic forms for what he wishes to express”. Again, neither method can be exclusive, since speaker and hearer are just as complementary as function and form, though not exactly in the same manner: the speaker makes a range of decisions, some more conscious and some less so, about how to correlate form, meaning, and function, and expects the hearer to make reasonably corresponding correlations. But in linguistic analysis, formal methods abstract away from the decision-making process and start with sample sentences already transcribed. If these have been invented for the purpose of analysis and description, the analyst takes on the speaker's role as well but from a non-representative standpoint of reduced functionality.

1.22 In sum, the heritage of Saussurian linguistics has been a set of interconnected structuralist and formalist theses about the nature of language and its description. The “langue” (language in abstraction) has been generally seen as a formal scheme of subsystems of levels, each internally constituted by formal arrays of differences. By implication, functional and intersystemic relations would belong to the sphere of “parole” (language in use). Functional linguistics would then face the options either of turning to “parole”, as Skalicka did, or of rejecting the “langue-parole” dichotomy entirely, as Trnka and Halliday did.{18} The most significant step would be to replace the dichotomy with a continual productive dialectic between “langue” and “parole”, between virtual and actual (cf. 2.10,13, 21,23, 29; 3.2, 15, 44; 4.16). It would be obtuse to imagine that every correlation between form and function entails an exactly corresponding movement between langue and parole; the correlation is feasible only because the two are alternative perspectives of the same phenomena engendered within an indissolvable dialectic.

1.23 Here we should consider carefully what sense we should attribute to the term “dialectic”. The classic sense of defining something in terms of what it is not might be confused with the static Saussurian conception of “difference” (culminating in the binary opposition) as the constitutive principle of order within the abstract system of differences (1.19). A more dynamic and productive sense would be that the function of an entity (not necessarily the entity itself) is defined by the contribution it makes to and receives from the entities which it is not but which are relevant to its occurrence. To see form-function correlations as dialectical in this sense is to assert on the one hand that form and function are different aspects and to insist on the other hand that the tension of their differentness is productive of significance. Applied to pairs like “langue” and “parole”, or “competence” and “performance”, or “deep structure” and “surface structure”, this sense affirms the alterity of two perspectives on a given data set yet sees each perspective as mutually constitutive of the other, including a systematic potential for intervention and change. Though we may not be able to perceive the same phenomenon simultaneously from both perspectives, we can continually shift from one to another and reappropriate the phenomenon.

1.24 The more formal versions of structuralism since Saussure, including generativism since Chomsky, have not proceeded as Mathesius imagined: they have taken form not merely as “the starting point”, but as the stopping point too. Theories and models have been obliged to overload and complicate formal aspects with implicit or disguised functional and semantic aspects. The multifunctionality and semantic openness of many forms and form-classes are in turn viewed as factors for theories and models not just to record and describe, but to resolve and eliminate. This demand can lead to such powerful reductions, idealizations, and exclusions that the completeness and relevance of the resulting description are difficult to establish. This dilemma may well be responsible for many of the disputes and crises we have seen in modern linguistics up to this very day.

1.25 To resolve the dilemma, the terms “structure”, “structuralism”, “function”, and “functionalism” should be reassessed within a more strategic and integrative framework. For a domain as complex as language, such terms should net be restricted in principle to specific aspects which seem easiest to appropriate, nor should the degree of formality and determinacy be determined a priori. The major emphasis should be placed on the activities of building structures and assigning functions, rather than on the labelled notations or classifications of structures and functions via formal criteria after the fact. These precepts are of course easier to formulate than to implement. But as long as they are discounted or postponed, linguistics will continue to be endangered by internal controversies and unrealistic demands. 

2. “Functional perspective” and the “sentence”

 2.1 Most functionalists today would agree with Halliday (cit. note 15, 46) that “a functional theory is a theory of language and is an essential aspect of any theory that attempts to explain the nature of language”. They would also agree with Danes and Vachek that the “mutual relations of subsystems, far from splitting the unity of the language system, rather strengthen and underline it”.{19}

2.2 The question is then how we implement these principles when we gather and organize our data. A task sequence we actually find in linguistic theory and research has included three steps: first to describe the structures of a set of sentences or utterances, second to describe their structural relatedness, and third to describe their strategic communicative potential. The first step has been taken by structuralist grammars, which discover the organization of a sentence or utterance by segmenting and labelling constituents. The second step has been taken by generative grammars, which postulate rules to make explicit the structural relations among the entire set of alternative sentences of a language (e.g. as “transformations”). The third step can best be taken by functional approaches, which consider how each alternative might be strategic for certain communicative intentions and decisions. However, considerations raised in section 1 indicate that these steps should be taken in parallel rather than (as they often were) in a detached or even competitive sequence.

2.3 The importance of this third step became clear to me when I encountered research on “functional sentence perspective” (hereafter “FSP”). The facts brought forth regarding FSP seemed to me unquestionably “linguistic” in any reasonable sense of the term. Indeed, the linguistic organization of sentences or utterances in a language with flexible word-order like Czech would seem quite unsystemic or unsystematic if FSP were disregarded. Even in a language with a restricted word-order like English, FSP helps to account for intuitively plausible differences between alternative organizations of the same linguistic material in a sentence or utterance. For Halliday, therefore, FSP is “part of general linguistic theory” and “a universal phenomenon” -- “a semiotic system without FSP would not be a language”.{20} This “universality” rests not on units, positions, and boundaries, but on “the cooperation of means of FSP, varying from language to language in accordance with the differences in grammatical structure”.{21}

2.4 I realized that FSP research constitutes a qualitatively distinctive dimension in linguistics. Its heritage is not so much a centralized “standard theory”, but a wide-ranging array of deliberations on perspectives and conditions. The leit-motif of this array is the intent to examine form, function, and meaning not each by itself, but in their mutual interactions both in actual communication and in contemplations or investigations of language. The description of such interactions is necessarily complex, but hopefully in the positive sense of providing a more encompassing and dynamic understanding of the organization of language (1.17).

2.5 On a very general plane, the “functions” identified in the Viennese tradition by Buhler and Kainz and extended by Mukarovsky and Jakobson (cf. 3.40) are relevant to Czechoslovakian research and were indeed conceived “in direct collaboration” with it, as I learn from Horalek.{22} “Buhler began from the conception of language as an instrument and thus continued the tradition going back to Kratylos and Plato” (op. cit., 41). For the speaker, the “expressive” function dominates (in Buhler's work, first “Kundgabe”, later “Ausdruck”). For the listener, the “appelative” function dominates (in Buhler's work, first “Auslosung”, then “Appell”). For the message itself, the “representative” function dominates (in Buhler's work, “Darstellung”).

2.6 This scheme of functions, however, is too general to offer more than a starting point for the analysis of utterances. Every utterance entails all three functions, and the modes of evidence for this entailment can be quite complex and diverse. Discourse often constructs or negotiates the world rather than merely “representing” it (3.13f). The speaker's “appeal” is to enlist the hearer in this construction and negotiation, not just to represent something, but to confirm or revise a presumed expectation, or to appear knowledgeable, reliable, or deserving of solidarity, and so on. All these motives can -have characteristic functional consequences.

2.7 The original Czech term introduced by Mathesius,{23} “actualni cleneni vetne”, might be translated “current sentence patterning”, but this version seems unduly narrow and form-oriented. The translation “functional sentence perspective”, following Mathesius' own German rendition “funktionale Satzperspektive”, has important advantages. As argued in section 1, functional, the first item in the term, implies a central concern for language patterns as interactions of form, function, and meaning. The third item, perspective, has the advantage of invoking not the pattern itself, but the viewpoint brought to bear on the pattern. A “perspective” is a wholistic cognitive disposition rather than an object, sequence, or unit; and resists being segmented into “minimal units” or “distinctive features”, or confined inside one level, or reduced to a formalism or formula. Moreover, the constructional and negotiational quality of discourse is kept in the foreground: the perspective constitutes as well as reflects and organizes the message.

2.8 The middle item in Mathesius term, sentence, has proved to be a central point of discussion. The main reason perhaps is that, as Danes observed,{24} “the sentence” has been “the kernel concept” in “systemic research”, yet “its content seems to be very variable, vague, complex, and undifferentiated”. It is not even clear where the sentence belongs in theory or method. For some authorities, like Saussure and Benveniste,{25} the sentence belongs to the domain of “parole”. For others, like Mathesius{26} and Chomsky, the sentence is the prime unit of the language system, while the “utterance” is the unit of “parole”. But, as Firbas warned,{27} this thesis should not lead us too readily to “the conclusion that non-utterance phenomena belong to the sphere of 'langue', whereas utterance phenomena belong to the sphere of parole”.

2.9 Due to its formal orientation portrayed in section 1, American linguistics was particularly concerned to situate the sentence inside the language system: first in structuralism as the sole “independent form not included in any larger” one{28} and thus as the structural framework in which other units should be discovered; then in generativism as the axiomatic framework for relating any possible formal arrangement of the language to the whole set of such arrangements through a rule system representing the speaker's “competence” (cf. 1.13; 2.2). Both the structuralist and generative notions of the sentence, however, remained uneasily positioned between “langue” and “parole”, and hardly in any dialectical sense. The segmentational method of structuralist analysis typically began with the sentence and worked down to the minimal units like phonemes and morphemes, whose abstractness and systemic status seemed decidedly clearer than that of the sentence itself. Being close to the utterance, the sentence might appear to be an actualised sequence of constituents belonging to the virtual system. The possible contradiction here -- the whole having a different status than the parts -- could be offset by the distinction between “type” (virtual system) and “token” (actual occurrence), which, like many other conceptions, was clear enough for phonemes (systemic unit versus produced sound) but much less so for sentences. In the generative method, the “type” status of the sentence was based on its rule-governed convertibility into other sentences of the systemic repertory, and the “token” status covered everything else, including the conditions of its production and reception. The later division between “deep structure” and “surface structure” ran along parallel lines, while the type-token problem in effect became an issue of formal conversion and notation (cf. 1.4, 2.19, 31).

2.10 The uneasy position of the sentence is a natural consequence of its status as both a functional and a formal unit and of the widespread uncertainty about whether and how far functional aspects belong to “langue” or (in generativism) to “competence” (1.21). In structuralism, these aspects were often set apart as “stylistic” or “rhetorical” issues rather than strictly “linguistic” ones. Generativism first allotted such issues to “surface structure” and later, when the potential of FSP for simplifying the description of word order was recognized (especially for languages other than English), to “deep structure”, the solution endorsed in Czechoslovakia by the Sgall group among others.{29} However, the argument in section 1 indicates that functional issues don't properly belong on any level, but arise from the dialectical interaction among levels, e.g. between deep and surface structure (1.22f).

2.11 Mathesius' solution was to “distinguish between the sentence as a pattern belonging to the language system and the sentence as part of the context, i.e. utterance, a component of the discourse” (cit. note 27). Definitions of the “sentence” in Czechoslovakian research have accordingly assumed a less abstract and formalizing character than elsewhere, e.g. “the sentence” is “an elementary verbal act of taking a standpoint toward some reality” (Vachek); or “the sentence” “is a field of relations, chiefly grammatical and semantic, functioning according to a given degree of contextual dependence, in a certain kind of perspective” (Firbas); or again “the sentence” of “a natural language” is “a systemic form of an elementary communicative linguistic act” (Sgall et al.).{30}

2.12 To consolidate their wider views of the sentence, the Czechoslovakian functionalists diverged from the structuralist and generativist intent to confine the sentence within “syntax” or “grammar” as a single “level” or “component” of language description. Instead, they undertook to distribute the sentence over formal, semantic, and functional domains. Following the functionalists' “fundamental thesis” that “language [la langue] is a system of means of expression that are appropriate to a goal” (1.1), Danes formulated “the utterance function” “of language systems: to be available for producing utterances” (cit. note 2, 127). “The stratificational hierarchy of language systems reflects the fact that its items are placed in different distances” from this “global function”: “the items on the lowest stratum” “exert this function” “most indirectly”, whereas “the items of the sentence stratum” do so “immediately”. Danes suggests that “the units of a lower stratum are less complex than those oft the higher”, while “the units of a higher stratum are constructed of units of the lower one, though not only of them” (op. cit., 128). “Consequently, the relation between a stratum and the next higher stratum” “may be interpreted as a means-end relation: the lower being the “means” and “the higher” being “the domain of functional implementation”. “This inter-stratal functional nexus” is an “internal” “constructional function” in which “both form and meaning participate”, “in contradistinction to the external utterance function”. Each “stratum” is composed of “the class of units showing the same degree of complexity and having” “the same constructional function” (op. cit., 129) — plainly positive complexity in the sense of 1.17.

2.13 The best-known proposal for correlating “external” with “internal” was Danes' “three-level approach”: the “grammatical structure of sentence”, “the semantic structure of sentence”, and “the organization of utterance” (cit. note 24, 226ff). The third level, that of utterance, represents “the functional perspective in a strict sense” and, in Firbas' words, “makes it possible to understand how the semantic and the grammatical structure function in the very act of communication”.{31} But the crucial functionality, I suggested (2.10), lies not on any one level, but in the “dialectical” interaction of all three in the dynamic sense proposed in 1.22f.

2.14 Doubtless as a concession to structuralist and generative linguistics, Danes suggested that “a strict differentiation” between the “semantic and grammatical levels” is “indispensable” for the “next step” of “ascertaining their systemic interaction” (op. cit., 226). But his citation of Firbas indicates, on the contrary, that a strict separation of levels would make it impossible to understand how either level functions even on its own terms. The only feasible order of the steps could be: first the inter-level perception and comprehension of the utterance and then the attempt at a “strict differentiation”. Even so, the considerations raised in section 1 indicate that the “strictness” of the differentiation might be more apparent than real, the product of an artificial and non-dialectical suppression of relevant interactions. And we might be entrained in a laborious and redundant reconstruction of parallel criteria and patterns for each level under different terms and thus incur a steep rise in “negative” complexity in the sense of 1.17.

2.15 Danes proposed to “distinguish” “the sentence” (1) as “a singular and individual” “utterance-event”, (2) as a “minimal communicative unit (utterance) of a given language”, and.(3) “as an abstract structure or configuration, i.e. a pattern of distinctive features” within the “grammatical system of a given language” (op. cit., 229). The first step would begin with the “utterance-event” pertaining to “speech (la parole)” and being “immediately accessible to our observation”. In the second step, “if we deprive such an event, by way of abstraction, of all accidental, singular, and individual elements”, we attain “an utterance which no longer belongs to speech” and has “non-grammatical but systemic means of organization such as word order” and “intonation”, including “many more features than only those belonging to the most abstract and general syntactic pattern of the grammatical system”. Here, the “sentence” figures as “one of all possible different minimal communicative units (utterances) of the given language”. This “utterance remains a part of context and situation”, “contains concrete lexical items and elements of modality” and “emphasis”. The third and “highest step of generalization” yields the “specific grammatical” “sentence pattern”, which “represents an abstract and static invariant structure (scheme), not a sequence of particular words” within an “utterance based on this underlying pattern” (op. cit., 230f).

2.16 Compared to the three levels, the three steps yield only a partial fit: it is fairly clear how the first and third correspond, but not the second step, whose job it is to discover how “the organization of utterance disposes of” “systemic means which have been, wrongly, classed with grammar, syntax or stylistics” (op. cit., 228). These means merit their own “theory of utterance” or “suprasyntactics” as “a special branch of linguistics” (op. cit., 230) (cf. 3.5). Presumably, the “semantic structure” belongs among the “non-grammatical but systemic means of organization”, but how it could be specifically isolated is not stipulated. Another key question here is what happens to the semantic structure during the third step: it would seem to be filtered out, but as suggested in section 1, it would still influence the discovery of “abstract structure”.

2.17 The “constitutive grammatical features” on the third level would include “parts of speech in morpho-syntactic classification, some morphological categories, and two relations of syntactic connexity, viz. dependence” and “adjunction” (op. cit., 231). The criteria are revealing: “adjunction” is “symmetric” (between elements of equal status), “transitive” carrying over to the other relations of each element), and “reflexive” (mutual), whereas “dependence” is “asymmetric”, “nontransitive”, and “irreflexive”. Of particular interest here are Danes’ stipulations that “adjunction” is an “a syntagmatic relation” and that “the word order belongs to the sentence pattern only” when it has “a grammatical function -- in Slavic languages very rarely” (loc. cit.). For many English grammarians, notably generative ones, word order is of course the primary carrier of grammatical function. In Czech, however, as Mathesius and his successors have commented, the word order follows FSP rather faithfully (cf. 2.28, 32; 3.43, 48). Danes soon recognized from his functional standpoint that the entirety of word order must have a broader and more diverse character, so that the more strictly “grammatical” “relations of syntactic connexity” should fall into a category by themselves. Qualifying one of his two syntactic relations as “asyntagmatic” because it does not entail a “dependence” between items, points to this recognition.

2.18 We see here an intriguing countermovement to the widespread tendency, described in section 1, to load and complicate the formal aspects or components and to construe word order per se as a dependency between elements seen from their next higher-ranking constituent (1.19, 24). Danes now proposed a distinction between the syntagmatic aspect proper (conjoining or adjoining of elements), which in Slavic languages is prominently handled with morphological resources, and the other aspects of word order, both concrete and abstract; and the term “syntax“ was expanded to cover all these aspects with equal regard. In comparison to Danes' concept ion, the “syntax“ of the generative school was both too broad in absorbing all of word order directly into the syntagmatic, and too narrow in excluding semantic and functional motivations for arranging words in one way rather than another. Danes was moving to integrate within a multi-level “syntax” the whole range of issues which western linguistics, especially in the U.S., has often marginalized or scattered among or between narrow levels of description, a rare exception being Peter Hartmann's treatise on Syntax und Bedeutung (Assen, 1964).

2.19 Danes' scheme accordingly transfers away from the “third specifically grammatical level” everything that can be accounted for in functional and semantic terms. He easily saw that this level could subsume not the sentence, but only the “abstract and static structure or scheme” of the “underlying pattern”. This substrate resembled the generativists' “deep structure” (introduced, I should point out, after Danes had formulated his “three level approach”), but the non-functional orientation of the generativists forced them to undermine the autonomy and centrality they attributed to the syntactic component by introducing semantic and functional material in disguise and by handling the notation in ways that clouded the relations between sentence pattern, sentence, and utterance. In Chomsky's Aspects, for example, “no careful distinction” was maintained “between the basic string and sentence itself”, and the discussion proceeded on the “simplifying and contrary to fact assumption that the underlying basic string is the sentence”, and the “base phrase marker is the surface structure as well as the deep structure” (cit. note 29, 18). Moreover, Chomsky said his “grammar does not, in itself, provide any sensible procedure for finding a deep structure of a given sentence” (op. cit., 141).

2.20 For Danes, the “sentence pattern” is “a syntactic structure” that “converts a sequence of words into a minimal communicative unit even outside the framework of connected discourse” (cit. note 24, page 230). From here, he derived the “condition” for “discovering sentence patterns” that the “corpus” be “the set of utterances” “employing the communicative function even outside context and situation”. Danes thereby arrived by an original route at a crucial problem in linguistic method at large. Whereas the apparent functional cohesiveness (or “connexity”) of sentence units has always been the practical basis for discovering them, he now proposed to make it the theoretical basis as well. Admittedly, I don't see how “a pattern” could have a “communicative function outside context and situation”. Linguists who imagine they are treating sentences out of context are typically replacing the real, integrative, and determined context with a fictional, isolative and non-determined context (cf. 1.5, 10). The sentence is extremely easy to use when constructing the latter type of context because sentence patterns are such a convenient and common organizational resource, not because they are a primary, self-sufficient substrate.

2.21 We might therefore reformulate Danes' “discovery condition”: the sentence is a multifunctional structure wherein the sentence pattern makes the syntactic contribution of co-organizing, through dependency and adjunction. a sequence of words into integrative constituents, the top-level constituent extending across the entire pattern. The sentence pattern thus coincides (is co-terminous) with the sentence but is not identical with it because the sentence is also dialectically constituted by semantic and functional contributions. Without these, the sentence would disappear and leave us with only an empty shell or outline lacking the crucial integrative potential mistakenly attributed to syntax alone. I believe this reformulation is fully in the spirit of both theory and practice of Czechoslovakian functionalism, as attested by such definitions of sentence as those cited above from Vachek, Firbas, and the Sgall group (2.11), but has been impeded by the influence of isolative and segmentational notions of “level” and “syntax” suggesting that the syntactic contribution remains intact when the others have been discounted. What would in fact remain is a non-linguistic formula which can be “converted” into a sentence only with all three contributions, whether by a linguist or anyone else; the syntactic one cannot manage this “conversion” by itself.

2.22 Recalcitrant problems with purely syntactic notions of the sentence are predictably most acute for elements not integrated within the more conspicuous structures and dependencies. In English, familiar instances are the “free adverbials” like unfortunately or in my opinion, “connectives” like consequently or on the other hand, and “interjections” like well and oh. Such an element does not depend on or “modify” another element in the rest of the sentence in a way comparable, say, to the relation between an article and its noun. The reformulation in 2.21 indicates why: the elements belong to the sentence not on syntactic grounds, but on functional grounds. A formalist approach is thus impelled either to exclude these elements or to situate them in “underlying” formal dependencies (e.g. inside an “embedding” within “deep structure”). A functionalist approach defines the notion of “constituency” primarily according to functional units and relations, and treats the formal ones as a contributing rather than constitutive factor. The notorious dilemma of the “sentence boundary”, inherent wherever spontaneous discourse is transcribed for purposes of syntactic analysis, thereby loses its virulence, because functional analysis does not insist on constrictive, deterministic description (1.5, 2.30). A “sentence” is a stretch of discourse which participants functionally treat as a sentence, whether or not the “well-formedness criteria” of a purely formal grammar can be directly applied. The latitude of this definition corresponds to the opportunities for different functional treatments in actual discourse.

2.23 The resistance among Czechoslovakian linguists against the notion of “competence” propagated by Chomsky's Aspects is specifically motivated by the insight that competence cannot be dominated by grammar, but must consist of the ability to organize multifunctional units and sequences, particularly sentences. The distinction between “grammatical and ungrammatical” thereby declines in favour of that between “functional” and “non-functional”. The competent speaker-hearer first and foremost performs a functional assessment of any given utterance, using the grammatical or syntactic contribution to support this assessment as a provisional but consensual indicator of relatedness. The phenomena of FSP accordingly document some of the strategies of this primary functional assessment and are central to “competence”, not peripheral or outside it, and the gap vis-a-vis “performance” is replaced with a continual dialectic (cf. 1.22f).

2.24 The semantic contribution of the “sentence” was emphasized in the spirit of “the Prague group”, who have “always striven to be free from the anti-semantic, purely formalistic bias” found among “some other linguistic groups of modern times” (Danes and Vachek, cit. note 19, 21). All the same, formulations were often rather cautious. “In regard to the relation between form and semantic content”, Firbas wrote (cit. note 21,40), it is obvious that not every semantic item is given its own form”. “Semantic relationships” in “natural languages” as opposed to those in “formal logic”, Sgall wrote,{32} are “not monovalent on both sides”, but admit of “asymmetric dualism”, as in “homonymy and synonymy”.

2.25 A more radical formulation of E. Benes raised “the possibility of distinguishing a specific dependence of word order on semantics”, “different from its dependence on grammar” and “based on the assumption: word order is determined by the combinations of abstract semantic categories” -- perhaps what Firbas meant with “the line of semantic structure”.{33} “In the act of communication itself, the assignment of names to the separate elements of reality and their arrangement into a particular sequence with a particular meaning” “may even precede their syntactical arrangement”. “The semantically desirable sequence of sentence components would be constituted first, a suitable grammatical-syntactic arrangement being chosen only afterwards” (Benes, loc. cit.).

2.26 Benes recalls Dokulil and Danes' notion of “a static content pattern” involving “relationships” (often called “logical”, cf. 3.22; 2.37f) like “agent and the action itself, bearer of the state and the state”, and envisions a “pattern of the semantic sentence structure” with “two aspects, i.e. as a means for the actual arrangement of the content elements and as a device for including the utterance in a context”.{34} But he offers a “simpler hypothesis”: “a particular arrangement of semantic categories in the static pattern of a sentence” “would result from the very nature of the discourse as a phenomenon bound to the passing of time, from the necessity to convert a simultaneously and multidimensionally existing extra-lingual reality by means of semantic categories into a linear sequence” (loc. cit.). Some “languages”, among them “Czech and other Slavonic languages”, might have “a semantic word order as an established pattern for the sequence of certain combinations of semantic categories”; “in the dynamic composition of the utterance the required pattern is either chosen ready-made or is modified according to the needs of the context” (cf. 3.8). Conversely, if we “eliminate the context”, we would have Ieft over those “word order configurations” “which result from the relationships of the semantic components” and of “content ideas reflecting extra-lingual reality” (op. cit., 271) (cf. 2.11; 3.13f, 16). Firbas similarly suggests a “semantic structure operating within that section of the sentence that has remained unaffected by, independent of, the preceding context”.{35} Benes goes further still by reserving the term “semantic” for “word order” that is “not affected by the context” (op. cit., 268). Yet this move encroaches on Danes' discovery condition for “sentence pattern”, which I reformulated a moment ago (2.20f).

2.27 One semantic scheme was the Sgall group's “systematic ordering of participants” centered on the “verb”: “actor, time, place, manner, instrument, dative, object of the type ‘what about’, objective (patiens), direction, objective complement, condition, purpose, cause” (cit. note 29, 67f). Many of these are familiar from traditional grammars with case systems. But a novel hypothesis is added, namely that these “participants”, in exactly the sequence quoted, form “a scale of communicative importance”. This scale may be implemented via FSP but may also be rearranged (cf. 3.9ff). Compare Firbas: “on account of their specific semantic character”, certain “words” may “weaken or strengthen the positions in the sentences in a more or less invariable way”.{36}

2.28 A multi-level approach to the arrangement of the sentences is naturally attractive for Czech scholars because it applies so aptly to their native language. Indeed, the conditions of Czech word order may well have been a critical factor motivating the whole enterprise of FSP and inspiring a new outlook on English as well. The old adage that English is best described by speakers of other languages seems quite fitting here.

2.29 Still, even Danes did not seem to have realized (witness his call for a “strict differentiation of semantic and grammatical”, 2.14) the full force and extent of the new approach. He was moving toward a genuinely dialectical conception (in the sense of 1.23) of “level” that departed fundamentally from the established descriptive conceptions wherein cross-relations between levels were invoked only for constituency, e.g. between the levels of “morpheme”, “word”, and “phrase” (1.19; 2.18). This tactic consolidated the form-based levels in a hierarchy of size, but marginalized their semantic and functional interactions. The Czechoslovakian approach, in contrast, disavowed the primacy of form and asserted that the inter-systemic or inter-level relations were the decisive ones because only they could indicate the functions of elements. Danes pointedly demonstrated this assertion with his multi-level functional “syntax” intended to unite form with function and meaning at every step; he reduced the “syntagmatic” to the domain of strict “dependences” and redistributed “word order” across three domains, each of them mutually presupposing and presupposed by the other two.

2.30 What this approach meant in principle might well not be immediately clear, particularly to Western linguists who would see only a confusion of their own tidy pigeonholes and categories. In practice, as the FSP literature continually reveals, it meant that any description of the form of a sentence could only be part of a. more complex scrutiny of communicative functions. Just how far such a scrutiny should proceed would be naturally hard to determine. Linguists accustomed to constrictive formal analysis may be unsettled by expansive functional analysis (1.5, 2.22). Indeed, there can be no strict a priori criteria for the extent or direction of relevant factors in a functional account (1.16, 25).

2.31 The danger was therefore imminent of misconstruing the functional scheme as a variant of prevailing formal-structural ones. Danes' first step of starting with the “utterance-event” “immediately accessible to our observation” (2.15) was often marginalized by the custom of adducing isolated, invented sentences, which are at best utterance events with a reduced functionality (cf. 1.5; 1.21). This custom jumps right into his second step, namely “depriving such a event, by way of abstraction, of all accidental, singular, and individual elements”, yet does not retain “systemic means of organization” whereby the “utterance remains a part of context and situation”. This second step is likely to be the most difficult and disputatious; in fact, competing linguistic methods differ here more than anywhere else. To evade the effects of context and situation, strictly formal methods leave the context-related part of this step entirely implicit and hurry onward to Danes' third step: extracting the “specific grammatical” “sentence pattern” which “represents an abstract and static invariant scheme”. The outcome of such tactics is that the relations between the “observed utterance-event” and the “abstract sentence pattern” remain obscure, as do all parallel relations: between parole and langue, between surface structure and deep structure, between token and type,(37) and so forth.

2.32 It would of course be premature to claim that functional approaches have precisely mapped out the second step such that it can be executed reliably and uniformly in every case. But the practice of functionalism has at least reinstated it firmly within linguistic inquiry. and demonstrated its importance, even -- or especially -- where disputes arise about individual points or claims. In early work, the leading role of FSP as principle of word order in Czech led Mathesius to decry English for being “insusceptible” or “heedless”, “frequently disregarding the requirements of FSP altogether”.{38} A major achievement of his successors has been to revise this negative assessment by revealing the multifunctional nature of FSP. English is not heedless, but, due to other constraints on word order, has a greater diversity of resources for FSP (cf. 1.10).

2.33 If the sentence is to be seen as a genuinely multifunctional unit, we must consider a wide range of factors influencing its organization. We need to transcend the traditional formal orientation looking only at units placed in certain positions and marked off by boundaries. A functional orientation can provide new insights into the selection and construction of such entities without being constrained to maneuver so narrowly along the borders of form (cf. 1.6, 8).

2.34 We can begin with Mathesius' portrayal of the sentence as a unit with a “point of orientation” (“vychodiste”) or “basis” (“zaklad”) and a “core” or “nucleus” (“jadro”). Although they have often been interpreted as such, these terms were not primarily positional or segmentational, i.e., were not meant to be names for elements or constituents of the sentence. Mathesius' original pair has been generally matched up with THEMA and RHEMA in Classical and Germanic studies and with THEME and RHEME in English studies. The pair TOPIC and COMMENT has been preferred by American linguists and their followers,{39}, who (as we might expect from section 1) have been the most determined to apply them to positions and segments, e.g., when Chomsky “defined the Topic” “as the leftmost NP [noun phrase] immediately dominated by S [Sentence] in surface structure and the Comment” as “the rest of the string” (cit. note 29, 23/68/70/163/230).

2.35 Some of these terms have appealing everyday senses that fit their technical senses only roughly. In popular parlance, “theme” and “topic” often refer to the subject matter or main idea of a discourse, and a clear counterpart term is lacking. “Comment” usually designates any remark indicating the speaker's own viewpoint about the subject matter, whereas “rheme” doesn't appear in everyday usage at all. A compromise might be in order: to select one term set for technical description and retain the rest in the commonsense usage. Since “theme” and “rheme” are evidently the most thoroughly explored terms in genuine functional research, I would favour (and have usually done so in my own writings) reserving them for FSP and using “topic” for general subject matter and “comment” for an incidental remark.

2.36 Both pairs of terms have been understandably associated with the familiar pair SUBJECT and PREDICATE. The latter are themselves multifunctional notions whose diffuse history points in at least three directions: grammar, logic, and psychology. In traditional descriptions, the standard practice has been to conflate the three, which seemed plausible enough as long as each example is so devised as to make them coincide. The coincidence allowed the ancient functional notion of a sentence as a unit in which something is made a “subject” and something is “predicated” about it --- compare Aristotles’ “hypokoimenon” (“the foundation”) and, “kategoroumenon” (“that said about it”), which he considered notions of “logic” -- to coexist alongside the formal notion of constituency between “noun phrase in subject position” and its “agreeing verb phrase in the predicate”.

2.37 Modern linguistics came to see distinct advantages in emphasizing non-coincidences. A famous instance, already cited (1.13), was Chomsky's much-touted observation that John is the grammatical subject in both (6) and (7), but the logical subject in (6) and the logical object in (7).{41}

(6) John is eager to please.

(7) John is easy to please.

His “transformational grammar” would derive (6) from John pleases someone and (7) from someone pleases John, these two being “kernel sentences” wherein grammatical and logical subjects once again coincide. In effect, such a grammar regards this coincidence as the most basic sentence formatting and all others as derived. A similar implication may apply to Chomsky's later notion of “deep structure”, which was said to be the home of “functional notions” like “logical Subject”, whereas “the so-called grammatical Subject” belonged to “surface structure”; here too, “a grammatical Subject may be a logical Object'“ (cit. note 29, 220f). Indeed, this kind of “difference” “provided the primary motivation and empirical justification for the theory of transformational grammar” (op. cit. 70).

2.38 Chomsky’s terms “logical Subject” and “Iogical Object” (the latter being a featured constituent of the “Iogical Predicate”) evidently designated the agent of an action and the entity affected by the action, respectively.{42} If a “logic” interpreted in its proper sense as “a ruIe system for determining new vaIues from known vaIues”, I see no strong reason to use the term “logical” here rather than “semantic”, the term preferred by many linguists anyway. Plausibly, Chomsky's appeals to “logic” were one tactic for including some semantic information in his grammar without designating it as such -- whence Danes' (cit. note 24, 226) verdict that “Chomsky apparently did not respect the difference between the grammatical and the semantic level in syntax”. Chomsky’s inclusion tactic resembled that of traditional grammar, which however made no claim to be “purely formal” (cf. 1.13).

2.39 The term “psychological subject” had a different tradition, due to the influence of nineteenth-century scholars like G. von der Gabelenz, H. Paul, and Ph. Wegener, and was carried over into the twentieth by the Czechoslovakian scholars J. Zubaty, V. Ertl, and F. Travnicek.{43} The “psychological subject” would be the idea the speaker has in mind as a starting point for creating a sentence, and the “predicate” would be the idea to be conveyed about this starting point. Such a notion resembles Mathesius' own conception (e.g. “vychodiste”) and a number of others we shall see later on. But the terms are now generally considered less appropriate or misleading, chiefly I think because “subject” and “predicate” are more pressingly needed for grammatical concerns.

2.40 The means of finding “subject”, “predicate”, and “object” did not seem problematic for traditional grammars with their orientation toward Latin, where we readily find a suitable nominative case, agreeing verb, and accusative case. The Latinate terms were reapplied to many modern languages, a practice which, with certain exceptions (e.g. Chinese and Japanese),{44} did not encounter large obstacles. A functional approach, however, cannot stop with the analysis of a sentence into “subject” and “predicate”, but must go on to inquire what functional criteria are involved in formatting the sentence and these two main constituents in this way rather than some other. Most linguists have no trouble taking a sample sentence and converting it into another grammatical format, e.g. active into passive. The new format may be a paraphrase of the old in terms of semantic content, but is very likely to differ in terms of FSP. Thus, an explicit and formal theory of paraphrase via reformatting, e.g. a “transformational grammar”, can only be one stage in an account of the functional relations between the alternative formats it describes (cf. 2.31). 

3. Thematic and rhematic as multifunctional aspects 

3.1 A productive way to survey FSP research might be to examine how sets of terms Can be aligned with the basic complementarity of thematic and rhematic. I have listed some prominent pairings in Table 1, with the left-hand term for thematic, and the middle term for the rhematic. The term in the right-hand column suggests the “dimension” along which the pair of terms might be situated. I should caution that some research focuses on just one such pair, while other research appeals to several pairs, often without specifying just how they agree or differ. Sometimes too, just one term of a pair is explicitly treated and the other is left implicit or mentioned only in passing.

3.2 The first pair, EARLIER versus LATER, is already open to a dual interpretation. The one favoured for a spoken sentence is the time dimension. Bolinger’s paper on “Linear Modification”, which was well received in Czechoslovakian research,{45} suggested that elements coming later are steadily more defined via those coming earlier. In his view, elements as they are progressively added one by one to form a sentence progressively limit the semantic range of all that has preceded; this causes beginning elements to have a wider semantic range than elements toward the end” (op. cit., li17). This progression seems reasonable, and adds another way to look at the “semantic structure” of the sentence besides those suggested by Dokulil, Danes, Firbas, Benes, and the Sgall group (cf. 2.13, 25ff). But we should consider recent cognitive findings that the operations of discourse production or comprehension need not be scheduled in the same time sequence as the positioning of elements, due to the use of global patterns like “schemas” (3.35). Moreover, the range of the beginning elements would usually be limited by those in preceding sentences within the same discourse. Still, Bolinger's notion of “semantic range” is a helpful reminder that meaning is at least partly indeterminate until it is actualised in communication and is, like FSP itself, a dialectic between virtual and actual (cf. 1.22f; 3.28).

3.3 The interpretation favoured for written language and in linguistics that relies on it is the position dimension. The spatial dimension of writing influences theoretical conceptualisation, e.g., when Benes distinguishes between “left-hand” and “right-hand contexts” (cit. note 33, 273). Some FSP researchers decided to define “theme” and “rheme” directly as positions, the theme always appearing before the rheme. This decision, adopted by Travnicek, Boost, and Halliday, but rejected by Firbas, Benes, Danes, Erben, and the Sgall group,{46} has at least two important implications. First, the selection of linguistic materials for the early stretch of a sentence is viewed as “thematizing”{47}, but the motives involved still require explanation. Second, the grammatical subject of the sentence is routinely expected to be thematic, and we must ask whether “non-thematic subjects are “peripheral” or whether they “are explainable by the same principles as thematic” (Firbas, cit. note 35, 240/253). Thus, the direct positional interpretation of “theme” and “rheme” does not by itself resolve the crucial issues and creates a fresh need for non-positional terms like “given” and “new” (cf. 3.24).

3.4 The second pair, FREQUENT versus RARE, seems simple enough -- we might just count elements and compare our totals. But numbers and statistics may not be helpful for so complex a phenomenon as language, where enumeration seldom constitutes explanation. No doubt the notion of “langue” should rest upon items and relations that occur frequently in “parole”; but I see no way to set a numerical cut-off point beyond which items cannot belong to “langue”. FSP compounds such problematics by diversifying the range of phenomena one might count.{48} We might define “thematic” elements as those which occur, either in form or meaning, more often throughout the discourse than the rest. Or we might assume we can intuitively identify thematic elements, and try to count how often they appear, say, in the position of grammatical' subject in a corpus of sentences. But the main task is still identifying and describlng the phenomena we decide to count. Like any formal or mathematical representations, statistics is productive only at the stage where we have a fairly reliable understanding of what we want to put into formal or quantitative terms (cf. 3.65f).

3.5 The third pair, STATIC versus DYNAMIC, is important because “the non-static dynamic character of language” is a “fundamental principle of the Prague School” (Danes, cit. note 2, 132). “It is exactly the emphasis laid on the dynamic nature of the system of language, together with its methodological implications and consequences, that might prove to be a contribution of some significance by the Prague group to the general context” of “language research” (Danes and Vachek, cit. note 19, 28). Moreover, “the designation 'functional sentence perspective’” itself serves to “appropriately indicate the active, ‘dynamic’ functioning of the semantic and grammatical sentence structures in the very act of communication” (Firbas, cit. note 38, 117). “What is for FSP of particular importance” is “the proper process of the dynamic building up of an utterance” and “of a supersentence discourse unit” (Benes, cit. note 33, 273). Danes' proposed “theory of utterance” (2.16) would cover “all that is concerned with the processual aspect of utterance in contrast to the : abstract and static character of the other two levels”, and would thus capture “the dynamism of relations between the meanings of individual lexical items in the process of progressive accumulation” and between “all other” “grammatical and semantic elements” (cit. note 24, 227). This “dynamism” “arises out of the semantic and formal tension and of expectation in the linear progression of the making-up of every utterance” (loc. cit.)