NEW FOUNDATIONS CH TWO

II. Toward a Science of Text and Discourse

II.A. Paradigms for the study of language

1. If we define a text as a communicative event (I.34), a discourse would be a set of interconnected texts, the primary instance being the conversation. It would follow that text and discourse are the main channels for people to encounter language. If so, all work for cultivating or studying language would have been at least implicitly or indirectly concerned with texts, even though only a few were explicitly or directly so. Before moving ahead, we can look back on past disciplines in this perspective.

2. Each discipline is circumscribed by a °schema° or °scenario° of shared concepts and methods regarding what questions to ask and where to seek the answers. The currently fashionable term is the paradigm, which the °philosophy of science° attributes to each normal science performing its routine work of solving problems (‘puzzles’). Periodically, a paradigm shift occurs, most radically in the scientific revolution. To cover its broad range of concerns, a °science of text and discourse° must ‘shift’ to a transdisciplinary paradigm that does not just displace some older paradigm with a new one, e.g., ‘sentence linguistics’ with ‘text linguistics’. Instead, ours should be a °meta-paradigm that integrates multiple paradigms° in order to situate their concepts and findings in wider contexts where disciplines can interact in full solidarity to develop a detailed °ecologist program for sustainable social progress° (I.60; II.132). To encompass the range of rich,interconnected issues we face today, we must transcend the philosophers’ popular but outmoded scenario of competing paradigms locked into a Darwinian struggle for survival (cf. III.24, 181; VII.199). In that spirit, it is essential to build upon our precursors and to appreciate the problems they faced and the solutions they attempted.

II.B. Textuality in grammar, rhetoric, and logic

3. The oldest and richest tradition belongs to the maintenance and interpretation of prestigious texts, typically official, legal, ritual, sacred, or poetic. Evidently, the text, particularly the written inscription, was widely esteemed a potent social instrument for expressing and transmitting the important knowledge of the culture. This esteem is most evident when prestigious texts are cited in legitimizing discourse about what is right or binding, witness these two examples from the 15th and 20th centuries [8-9] (i.a.).

[8] I am talking of the statutes that have the force of canons and of decretals that are universally binding on the church […] since it is useful to introduce any reform on the basis of precedent, I would submit an imperial letter which is credited to Constantine (Nicholas of Cusa, ‘On the supremacy of general councils in church and empire’, 1433)

[9] [Prime Minister ] Begin reviewed the text in detail, and finally concluded, ‘This document is not a proper basis for negotiations’ […] He focused on words and their meaning […] In all, the American team prepared 23 versions of the Framework for Peace. I wrote the original Sinai agreement personally, and there were eight different texts before we finished (Jimmy Carter, Keeping Faith, 1982)

Evidently, many people believe having written text settles things for good.

4. Yet the elite rank and specialized training of many experts on prestigious texts — say, for the Vedic priests or the Kabbalists of the Torah, or for today’s attorneys of corporate law — suggests that the text can be a highly problematic social instrument. Paradoxically, the experts apply specialized skills and procedures to construct a meaning that has an adaptive value for them while purporting to have only registered and presented the one correct or true meaning . The authority of the text is all too readily co-opted for authoritarian moves of power (cf. II.129; III.10; V.14). The °reifying ‘formalist’ myth of the self-sufficient text with a stable and determinate meaning°, which inflates the individual over the social (cf. I.39; II.110; III.107), serves the dual agenda of enhancing the prestige of texts and securing the rights of interpretation for elite groups (V.80; VII.199).

5. So in most historical settings, textual skills decide whether °access to knowledge promotes either power and inequality or else solidarity and equality°. Discoursal power moves are easy to °mystify as ‘literal’ and ‘true’ statements about ‘reality’° because few people properly appreciate the °potential of discourse to construct and negotiate reality° (cf. VIII.75, 131). Conversely, discourses of power are hard to °demystify° because doing so requires us to raise our critical consciousness and suspend our easy routines of understanding talk in order to draw wider and subtler connections (cf. I.59). We need to critique such familiar expressions as ‘nature’ and ‘reason’, e.g. in [10], and recognize such discourse to be a direct precursor of today’s myth of ‘intelligence’ and ‘aptitude’ being pre-determined by nature (VII.64).

[10] Every constitution is founded on natural law [that] exists by nature in reason. [the] wiser and more excellent are chosen as rulers in order that, endowed with a naturally clear reason, […] they may choose just laws and by these govern others and hear cases. […] thus those who are strong in reason are by nature masters and rulers of others (Nicholas of Cusa)

The mystification of discourses of power has been abetted by the disinterest among language-related disciplines in explaining the relevant strategies and providing an explicit theory for such a long-standing and important practice. Evidently, powerful people are not eager to have it made public how they use discourse in ways that are quite adaptive for them and quite maladaptive for the disempowered.

6. Three main disciplines have traditionally been concerned with language. The discipline of grammar has sought to expound the organization of a language in terms of forms, patterns, and rules. Typically, °prestigious texts° were chosen as sources, which enhanced the prestige of grammar and encouraged qualified scholars to study and cultivate it. Using such sources placed traditional grammarians fairly close to ‘text linguists’ and kept them from anticipating the 20th‑century rift between ‘language study’ versus ‘text study’ (cf. II.24). Their own specialized education and ‘high literacy’ legitimized the early grammarians’ claim to ‘know the language’ and encouraged many of them to act as °language guardians° by adopting conservative, elitist standards and by making °prescriptions for ‘correct’ usage and proscriptions against ‘incorrect’ usage°. This strategy was highly adaptive in creating a lucrative and self-perpetuating market for expertise about grammar. In practice, the standards were largely constructed along social lines to accentuate the dual distinctions between writing versus speech and between a .i.prestigious variety or dialect versus one or more non-prestigious ones (cf. VII. 251-55). The reliability and generality of such grammars for representing actual usage remained uncertain, insofar as the things people were not supposed to say were precisely ones which many people did say.

7. The conservative quality of grammars was reinforced by the strategy of recycling earlier grammars, even ones from other languages, e.g., when the English abbot Ælfric Grammaticus (ca. 955-1020) relied in [11] on the Latin grammarians Priscian of Caesarea (5th century A.D.) and Aelius Donatus (4th

century). Like many later grammarians, Ælfric assumed that English grammar should be legitimized by borrowing upon Latin grammar and using the same descriptive terms. Fortunately, the models they admired, such as Priscian’s Institutio de arte grammatica and Donatus’ Grammatica urbis Romae, took a very broad view of grammar, witness Donatus’ sections on the ‘voice, letter, syllable, feet, tones, schemes, and tropes’ (de voce, de littera, de syllaba, de pedibus, de tonis, de schematibus, de tropis). Hence, Ælfric’s notion of ‘grammatica’ (Anglo-Saxon ‘stæfcræft’, ‘skill in letters [of the alphabet]’, but also ‘learning’, ‘study’) covered the sounds of speech and the construction of meaning as well, in each case foregrounding written language. In modern terms, the ‘grammar’ extended into ‘phonology’, ‘intonation’, and ‘semantics’ (cf. II.29f, 48ff, 63).

8. Rhetoric was a more directly textual discipline whose social function was to teach active and public skills, especially for oratory, rather than to cultivate the passive and private skills gleaned from the grammatical study of °prestigious texts°. The rhetoricians saw in language less a catalogue of ‘correct’ forms and patterns than an armory of discourse strategies for practical goals. This scope was not necessarily narrower than grammar, because the rhetoricians, while not trying to cover all the points a grammar might address, emphasized richer factors of context, e.g., how to persuade particular audiences. Rhetorical effectiv eness may not coincide with grammatical ‘correctness’; some audiences might see an inappropriate power move in the use of a cultivated style preferred by grammarians. Rhetoric (with the Greek root ‘eírein’ meaning ‘speak’) has therefore been more political and populist and centered on oral discourse, whereas grammar (with the Greek root ‘gramma’ meaning ‘letter’) has been more academic and elitist and centered on written discourse. Privileging grammar over rhetoric in education right up into modern times reflects the policy of granting preferential knowledge access to reward ‘high literacy’ and ‘standard’ dialect rather than the skills of popular communication (VII.162f, 169f, 212f).

9. Recognizing the public demand for training in rhetoric and the profits of the schools already offering it (e.g. the school of Isocrates), Aristotle expounded his famous Art of Rhetoric in search of a systematic philosophical and psychological foundation. His method of classifying into subtypes has proven vastly influential ever since, even among modern linguists (II.34; V.45). For example, he distinguished between the ‘deliberative’ rhetoric for reaching decisions in assemblies, the ‘forensic’ (or ‘litigious’) for arguing cases in court, and the ‘epideictic’ for displaying elegant or ornate language. His advice about the proper style or usage of language was ambivalent, due his admiration for poetry and stage drama (e.g. by the sonorous Euripides). He recommended ‘appropriate’ language and the ‘normal idiom’ but also ‘exotic’ and ‘distant’ language that is ‘artificial’ but seems ‘natural’. For writing, his ominous suggestion that content ranks below style has stirred considerable annoyance (cf. V.39).

[12] the virtue of style is to be clear, and to be neither low nor endlessly sublime, but appropriate . […] the discourse must be made to sound exotic, for men are admirers of what is distant, and what is admired is pleasant. […] one must do this without being noticed and give the impression of speaking not artificially but naturally. […] the technique is well concealed by drawing words from the normal idiom, […] we should make little use of exotic, compound, and artificial ones; […] they involve too great a change from what is appropriate. […] written speeches have more effect through their style than through their intellectual content.

Despite a brief plea for ‘correct’ grammar (agreement of Gender and Number), he differed prominently from the grammarians (and even more from the logicians) by describing how to build rational or ‘logical’ arguments and ‘proofs’ upon uncertain or subjective premises, and how to gain power by exploiting the audience’s emotions (anger, contempt, gentleness, friendship, love, pity, envy, jealousy). He classified people into ‘character types’ by age and by social rank, noting that ‘the rich are arrogant and overbearing’ and ‘mindlessly happy’, for ‘their wealth is the highest measure of all things’, and ‘they believe themselves privileged to rule’; — just what the wealthy backers of today’s °New Right coalition° devoutly and mindlessly believe (cf. VII.32).

10. Logic was a discipline aspiring to complete the search for a universal system of knowing in the discipline of ‘philosophy’ (from Greek roots, ‘love of knowledge’). Whereas some philosophers invoked an absolute ‘first principle’ such as ‘Nature’ or ‘God’, others turned to language as the privileged mode for knowing, meaning, defining, asserting, and so on (whence ‘logic’ from the Greek root ‘logos’ meaning ‘word’ or ‘speech’ and, technically, ‘sentence’). Early logic, with strong affinities to mathematics and geometry, sought to construct proofs on certain and objective premises rather than on the uncertain and subjective premises of rhetoric. So whereas the grammarians cultivated a body of prestigious texts, and the rhetoricians composed tactical arguments for debates, the logicians sought a system of principles (or ‘axioms’) whereby statements and arguments could be constructed and proven true or false, independently of text type and context, or of speaker and hearer and their beliefs, attitudes, or goals. The term ‘logic’ might therefore justly subsume all relationships and methods whereby the value of a term can be strictly determined through formal steps from the known value of other terms (cf. III.167f). In the idealized complete system, every fact, premise, and conclusion would be either true or false, freed from the rich and evolving contexts of everyday experience and ordinary discourse.

11. A favorite illustration has been the classical syllogism:

[13] major premise: All men are mortal.

minor premise: Socrates is a man.

conclusion: Therefore, Socrates is mortal.

Yet even this threadbare syllogism is not so simple and certain. The conclusion is valid only from a sparse reading of the major and minor premises by handling the words and phrases as mere counters or placeholders, but is convincing from a rich reading based on our knowledge of what the words mean and how the world is organized. The validity holds whether or not we know who Socrates was or what is meant by ‘man’ and ‘mortal’, because the syllogism merely uses language to restate the truism of set theory that a property holding for all members of a set also holds for each individual member (cf. II.50; V.113). But we are convinced because we do know all that, and we might be upset if the wording were changed to ‘all men are moral’ or ‘Xanthippe is a man’, even though the syllogism would still hold. We could only dispute the ‘facts’, which the logic doesn’t claim to supply. Nor is the logic relevant if somebody retorts that Socrates was ‘immortal’ because his life and works live on, or interprets ‘man’ in an unusually rich sense, as when Antony says of Brutus ‘This was a man’ because ‘the elements’ were ‘so mixed in him’ (Julius Caesar, V, v, 72-75). Logic disdains mixed elements.

12. Logic symptomatically underrates the essential difference between language versus mathematics and geometry, whose exactness and universality hinge on being disconnected from content and context. The ‘meaning’ of numbers and their combinations is precisely the quantities they express; the ‘meaning’ of lines and figures is precisely their dimensional and positional relations (cf. III.158). These sparse conditions do not hold either for a ‘natural language’ like English or for human knowledge of the world and society. A disquieting trade-off impends: the more rigorous and precise the logic becomes, the sparser the meaning and the lower the cognitive and social relevance of its statements (cf. II.50, 57; III.168).

13. Still, over the centuries, the project for using logic to definitively capture truth and meaning has enthroned logic as the prime model for all human thought and reasoning (II.96; III.167), both in ‘idealism’ and in some radical versions of its archrival ‘realism’, such as ‘logical positivism’ holding that certain (‘positive’) knowledge is based on observable physical phenomena, and ‘verificationism’ holding that knowing the ‘meaning’ of statements equals knowing how to verify them by observing the reality they describe. Similarly, the science of language has at times used ‘formal language’ as the leading model for ‘natural language’ (cf. II.58, 99; III.167-73). So philosophy, language science, and psychology have projected their own procedures for designing formal models and representations over onto human processes of thought and language at large. I hold this projection to be a category mistake whereby a theory is ‘short-circuited’ onto the domain of practices it purports to describe or explain (cf. II.96; III.169, 180; V.124; VII.63; VIII.17). How far human thought or language might be formal or logical is a question to be empirically discovered and not just quietly pre-empted or taken on faith (II.79;III.173; IV.30).

4. Fig. II.1 sums up how grammar, rhetoric, and logic might be assessed as three approaches to

texts. Typically, a grammar has been derived from passive skills honed on a body of prestigious texts, mainly written; a rhetoric by practicing more active skills on oratory texts, mainly oral though sometimes transcribed and embellished; and a logic by formal skills with using notations and determining the validity and truth values of mainly artificial texts.

15. Despite their disparate perspectives on texts, grammar, rhetoric, and logic co-existed over the centuries in a ‘trivium’, a trinity of eminent scholastic domains (called ‘liberal arts’). Some early authorities saw no tension among them, e.g., when Aristotle sought to merge rhetoric with logic and treated grammar as well (II.9). Our ‘modern’ age has seen them split up; grammar has survived best, heavily represented in schooling at all levels (VII.K.2), whereas marginal areas are left to rhetoric in departments of speech or journalism and to logic in philosophy. The heritage of the trivium has been a studious respect for high standards of style, cogency, and clarity that education has appreciated without being properly equipped to impart the requisite skills in equality (Ch. VII). Perhaps a °science of text and discourse° can contribute to a productive reassessment of criteria for grammatical standards, rhetorical effectiv eness, or logical validity.

II.C. Textuality in philology

16. The notions of ‘language science’ or ‘modern linguistics’ owe much to the field of philology, which was established in the nineteenth and century, and whose chief concerns were: (a) historical connections among successive stages of a language; (b) comparative connections among languages from related groups or ‘families’; and (c) geographical connections among various ‘dialects’ of a language. All three investigations have gathered their main evidence from written text-artifacts and devoted scrupulous attention to problems and inconsistencies in transmission and transcription. Historical studies, particularly by the ‘Neo-Grammarians’, sought to formulate ‘laws’ of language change, but also to produce reliable textual editions (e.g. the Hildebrandslied, 1812), and to cultivate the language skills for studying and using these — whence such academic domains as ‘Old English’ and ‘ältere Germanistik’. Comparative studies, particularly those of the ‘Indo-European’ group, inquired which languages were related, and whether the relations were due to common ancestry or to other factors like migration and borrowing. Geographical studies, often labelled ‘dialectology’, focused on forms or pronunciations that varied from region to region in regular ways. This work pioneered thorough methods to describe dialects apart from the usual pejorative social attitudes (cf. II.6; VII.49, 171; VIII.96, 150).

17. One eminent success in philology was to describe the great ‘sound shift’ that split off from the other ‘Indo-European’ languages the ‘Germanic’ branch that would later produce ‘Anglo-Saxon’ and later still ‘English’. For example, the ‘unvoiced stops’ (where the air passage is stopped but the vocal cords are inactive, II.29) written as ‘p’, ‘t’, and ‘k’, got shifted to ‘fricatives’ (where the air passes with friction) written as ‘f’, ‘th’, and ‘h’, whence Latin ‘pater’, ‘tres’, and ‘centum’ versus English ‘father’, ‘three’, and ‘hundred’. A shift could be strikingly regular, affecting virtually all relevant words and even progressing to a precise geographical line; unaffected items are probably later borrowings or frozen proper names, ritual terms, and so on. So the ‘sound shift’ shed light on all three branches of philology: describing the history of languages and their common ancestry as well as the evolution of dialects. The discovery nicely fit the 19th-century concept of ‘natural laws’ reaching all across the sciences. Here already, the study of sound systems fostered an impressively tidy theory of language, and detached the language from the discourse practices of social agents (II.29; III.137).

18. Philologists also pioneered the technique of working with authentic data from a corpus of language samples; and they carefully described what they found people saying instead of °prescribing° what people should say in the manner of traditional grammar. The key question was which forms were attested, i.e., °textualized° in preserved inscriptions or transcriptions for reliable identification and collation. Here, the textualizing of forms provided the means not merely to determine their function and meaning — as for all language (I.41) — but to decide which forms merited scrutiny at all. Although ‘Indo-European’ is commonly held to be a scholarly reconstruction and an abstract system of forms rather than an actual language, some unrecorded language(s) resembling it must once have been textualized in order to develop and stabilize the formal regularities we can trace in its descendants. And the recovery of genuine texts in any early offshoot of Indo-European is a sensational event, witness the stir in our own century over the discovery of Hittite going back 3,000 years.

19. Recent research has raised again the prospect that, contrary to the standard view among philologists, that Indo-European may not be unrelated to other language families after all. Supportive evidence includes the ‘-m-’ in First Person forms and the ‘-t-’ in Second Person forms consistently appearing across the whole group called ‘Eurasiatic’ by Joseph Greenberg, including also Uralic , Altaic, Korean-Japanese-Ainu, Gilyak, Chukchi-Kamchatkan, and Eskimo-Aleut. Some linguists, such as Morris Swadesh, have even revived the prospect, once discounted by philologists, of a monogenesis whereby all the world’s languages descended from a single one. How they might have evolved into separate groups is being actively discussed in the wake of startling findings that populations with different language families also differ in their genetics. These findings converge with burgeoning explorations of the relationship between language and genetics.

20. Fig. II.2 shows how the branches of philology might be seen as three more approaches to texts.

The historical branch compiled and edited documents and inscriptions to trace the historical stages in a language. The comparative branch mapped the relations among languages or ‘language families’ indicated by patterns of attested forms complemented with painstakingly reconstructed forms. The geographical branch used surveys and questionnaires to elicit the dictation or transcription of regional varieties. In each of the three, the concept of ‘language’ varied to focus on evolution, comparison, and variation, respectively.

21. Philology grappled valiantly with the problems of analyzing authentic data and of weighing textual evidence against conflicting or missing evidence, especially when postulating non-attested forms. With living languages, in contrast, informants can be consulted, and the obstacle may be not too little evidence but too much — too many diverse opinions and too little consensus about the vocabulary, grammar, or sound systems in various dialects or geographical areas. Most techniques of taking dictation rarefy the factors not easily represented through the conventions of writing (III.195). And if schoolmasters are the informants, as for the monumental Deutscher Sprachatlas research, their usage may be more self-conscious than the usage of ordinary speakers.

22. Despite such problems, philology inaugurated major advances in the directions later pursued by modern linguistics. The main concern was no longer grammatical, rhetorical, or logical standards, but the systematic properties of all the data that could be assembled in a corpus. Philologists appreciated the °prestigious texts° of literary and poetic discourse and scrupulously collected and edited them. But they also worked to uncover and preserve any surviving samples of non-literary discourse, such as statutes, proclamations, proverbs, prayers, incantations, personal letters, and so on. This concern prefigured the more resolute move of linguistics to affirm the centrality of spoken ordinary language over written elite language (cf. II.24ff). Still, many modern linguists have not properly acknowledged their debt to philology. They may have been motivated by rivalry or by competition within academic language programs treating ancient and medieval languages from a mainly literary viewpoint. Yet these programs did pave the way for modern language programs in the schools and universities. Perhaps a °science of text and discourse° can requite our own debt to philologists by providing further perspectives for the study of the early texts they have so diligently preserved.

II.D. Textuality in modern linguistics

23. The term modern linguistics designates a field of research and method that emerged in the 19th century and was consolidated in the 20th. It resolved to be modern by making a fundamentally new departure from the earlier approaches to language outlined so far in this chapter. It would be scientific by framing explicit theories of language and implementing disciplined practices. It would describe what people do say rather than prescribing or proscribing what people ought or ought not to say. It would provide comprehensive coverage of the whole language rather than selective coverage of grammatical, rhetorical, or logical issues. And it would defend the primacy of spoken language, e.g. by developing elaborate methods for representing sounds more accurately than with conventional orthographies (especially for English, VI.56); although written language samples were often used and their visual clarity did some ‘cheap explanatory work’ in identifying units and their boundaries (cf. II.29, 34, 38; III.195, 202; IV.1)

24. However, the ratio between theory and practice in modern linguistics has remained profoundly problematic (cf. IV.21) The human practices connected to language are dauntingly vast and diverse, and certainly do not, by themselves, stipulate the form or the content a theory of language ought to have. So, many linguists — though by no means all, as we shall see — have attempted to stipulate a theory from the ‘top down’, i.e., in purely theoretical terms, which would stipulate the methods for connecting it to the practices of linguists and of speakers. A famous project of this kind was inaugurated by Ferdinand de Saussure’s ‘course in general linguistics’ (published in French in 1916), which declared the ‘true and unique object of linguistics’ to be language studied in and for itself’, termed ‘langue’ in French, and placed it in a stringent dichotomy against language use, termed ‘parole’. Henceforth, the ‘classical’ program for °mainstream linguistics° — as we find it reflected in influential works and in the agendas of university departments, professional journals and conferences, and so on — has typically constructed its theories of language according to tenets like these:

(a) Language is a phenomenon distinct from other domains of human knowledge or activity.

(b) Language should be described apart from the conditions under which speakers use it.

(d) Language should be considered a uniform, stable, and abstract system in a single stage of its evolution, i.e., in a ‘synchronic’ perspective at the present time rather than in an ‘evolutionary’ or ‘diachronic’ one across historical time.

(e) The description should be stated at a high degree of generality — what applies to the entire language or even ‘universally’ to all languages.

In the early stages, these tenets strategically suited the academic politics of establishing a °monodisciplinary normal science° in a sparse and self-conscious scientific climate. Disconnecting °language by itself° from issues in neighboring fields, such as literature, history, folklore, philosophy, psychology, and sociology, was expected to make the description more compact, unified, and rigorous by imposing strict limits on which data get accredited, i.e., accepted as worthy of investigation and placed into the established categories.

25. But a theory of this type raises a serious problem for practice: language is not encountered by itself but only in use. Undaunted, linguists undertook to construct a language as a theoretical ‘system’ presumed to ‘underlie’ all of its practical uses, though the preferred term was discover to uphold °classical realism° by implying that this system is a reality waiting to be found. The chief practices of discovery have been data-handling moves like these:

(1) collating : the data samples in a °corpus° are compared and contrasted to see what they have in common, e.g., which word types frequently co-occur;

(2) consulting informants : native speakers are asked to judge or rate data samples of their language, e.g., which among several versions of an utterance they would be more or less likely to say;

(3) generalizing : certain aspects of the data are construed to be general ones, e.g., that the ‘Subject-Verb-Object’ order of a sample set of English Sentences is a typical pattern for the whole language;

(4) rarefying : the ‘rich’ data as we find them in real discourse are rendered ‘sparse’ by disregarding certain aspects or details, e.g., variations in the actual pronunciation of the ‘same’ language sound;

(5) decontextualizing : the data are removed from the observed context and treated as if they could occur in isolation or in a wide range of contexts, e.g., irrespective of the social status of groups or speakers;

(6) introspecting : the linguists base estimations on their own °intuitions° about the language, e.g., which sentences conform to ‘grammatical rules’.

These practices were performed as °cognitive moves° while organizing materials and constructing theories and models, and as °discoursal and social moves ° while stating the results and informing colleagues through reports, conferences, journals, books, and so on. Because our own language strongly affects how we classify and label things, the discoursal aspects of ‘doing linguistics’ are far more influential and significant than has been widely recognized.

26. The practices for these data-handling moves differed between two broad approaches to language. Fieldwork linguists (as the term says) go ‘work’ in the ‘field’ of cultural and social activities and carefully record a corpus of °authentic data°: what native speakers of a previously undescribed language or dialect are actually observed to say. In contrast, homework linguists (to coin a new matching term) ‘work’ at ‘home’ (or in the office) with data that have been supplied from an outside source. The data may be authentic, as in a computerized corpus of real discourse (II.64). But they may also be °invented data° by the linguists, who are fluent speakers and claim to represent the whole community of native speakers. This tactic is highly problematic, and we need to consider why (II.36; IV.21).

27. In practice, fieldworkers never encounter °language by itself°, even if the official theory purports to describe it; you confront language in °rich connections with language use° and exploit modes of data that are not just ‘linguistic’, such as the procedures of buying and selling goods (II.33). Your work is always strongly °data-driven°, and °practice-driven°, especially when you join in the social practices of interaction and conversation. The practices supply continual tests: if your theory or your conclusions are wrong, you’ll get corrected, misunderstood, teased, or ignored. In contrast, homework can disconnect theory from practice by relying on your own introspection and intuition to formulate and describe what native speakers are presumed to know about the language by itself. The work becomes strongly °theory-driven°, especially when you pursue a construction that, by definition, does not directly manifest itself in data or in practice, and when you try to disconnect invented data from cognitive and social constraints . Nor are your theory and conclusions tested by being put into practice. Instead, theories and descriptions get evaluated mainly by standards of design, such as formality, rigor, elegance, or compactness (cf. II.43, 92). In exchange, authentic data such as everyday conversations may look messy or deviant, hardly worthy of attention (II.41; VII.133). So the quest to discover the language-system as an underlying reality can paradoxically lead radical homework linguists to replace the language with a °technical construction° which exists only in their own theorizing and can be connected to language data only when these have been removed from practice and translated into some °sparse notation°. Along the way, real speakers as informants get marginalized and silenced, and a written ‘standard’ variety parades as the language itself (VII.165; VIII.18) — the linguists’ version of the °fundamental contradiction between inclusive theory (e.g. for covering an ‘infinite set of sentences’) versus exclusive practice° (e.g. for disregarding ordinary conversations) (cf. I.6; II.41). And cultural memory gets erased to avoid questions about how concrete discoursal practices might evolve and for whose benefit (III.113).

28. After 70 or 80 years of research, the status of °language by itself° is still uncertain. Since it is not a manifestation but a theoretical construction, it cannot be verified by conventional observation of the practices of °language use°. So we can move to a higher plane and observe the progress of linguistics as a science . There, our testable hypothesis might be: if ‘language’ is indeed a °uniform, stable, and abstract system we can describe by ‘purely linguistic’ criteria°, then we should observe steady long-range increases on three ‘test scales’: (a) coverage, i.e., how much language data have been described; (b) convergence, i.e., how far various descriptions get corresponding results; and (c) consensus, i.e., how far linguists agree about how the description should be stated and assessed. What we actually observe, however, is not a steady overall increase but an uneven pattern of increases in some domains, and stagnation or short-range decreases in other domains. How far linguistics might count as a °normal science° (in the sense of II.2) is still not decided; its ‘normalcy’ fluctuates periodically and substantially (II.41).

29. We observe the stablest increases in the description of language sounds, which had already anchored the philologists’ concept of language being governed by ‘natural laws’ (II.17). In the domain of phonology (or ‘phonemics’), linguistics discovered a °uniform, stable, and abstract system° of phonemes: theoretical minimal units which correspond to the practical units of language sounds, and whose quantity and nature can be precisely described in practice for any language by the physical and mental criteria that ‘differentiate’ them (cf. II.45, 48). Physically, each phoneme is connected to articulatory events and locations, e.g., a ‘voiced dental stop’ such as /d/ produced when the vocal cords vibrate and the air flow is blocked by the teeth (cf. II.17); the concrete physiological processes are investigated in the kindred domain of phonetics. Mentally, each phoneme must be capable of differentiating between elements (words or word-parts) that also differ in meaning, e.g., /d/ versus /t/ in ‘hid’ versus ‘hit’ (III.92). These two sets of criteria nicely converge in practice to give tidy results: complete coverage and a high consensus among phonologists. The description also reaps some ‘cheap explanatory work’ from the visual match between the notation of many phonemes and written letters, or technically, the graphemes, of the popular Roman alphabet — even though spoken language was the official domain to be described (II.23).

30. This huge success on our three test scales made the study of sound systems in phonology into the model paradigm for early modern linguistics, whence the series of ‘‑eme’ terms (e.g., ‘morpheme’, ‘lexeme’, ‘syntagmeme’, ‘sememe’) modelled on ‘phoneme’. Henceforth, °mainstream theories° confidently projected language to be an array of °uniform, stable, and abstract° ‘subsystems’, usually called levels, each consisting of a repertory of theoretical minimal units which correspond to the practical units in the language data and which constitute a system of differences (cf. II.39, 45, 114, 123). A complete description of a language would be the sum of the descriptions for each subsystem, supplied by linguists investigating the several areas within a neat division of labor. Research of any scope or in any area could contribute on its own terms, e.g., by doing ‘pure phonology’ or ‘only syntax’. In this fashion, linguistics would be ’mono-disciplinary’ and would also contain a set of smaller ‘monodisciplines’.

31. Impressive success was also attained in the domain of morphology (or ‘morphemics’). Here, the theoretical minimal units are the meaningful forms called morphemes, which correspond to the practical units of word-parts, such as word-stem, prefix, or suffix, and of non-segmentable words. This domain offered vital support for °fieldwork in ‘Peripheral’ regions on previously undescribed languages of lesser diffusion° that often °defamiliarize the Western fieldworker° by presenting much more elaborated systems of meaningful forms than do the familiar ‘Western’ languages of ‘Center’ regions (e.g. English and French) (cf. II. 37, 86; VIII.9). The ‘polysynthetic’ languages spoken, say, by Native Americans join morphemes into long word-like constructions that would be expressed by complex phrases in other languages, e.g., [14] from Paiute of south-western Utah (reported by Edward Sapir), or even by whole utterances, e.g., [15] from Aymara of Peru and Bolivia (reported by Martha J. Hardman) (cf. II.62).

[14] wii - to - kuchum-punku-rügani-yugwi - va - ntü - m(ü)

knife-black-buffalo - pet - cut up-sit (plural)-future-participle-animate plural

‘they who are going to sit and cut up with a knife a black cow (or bull)’

[15] aru-si - kipa - si - p - xa - ña - naka -sa - ki - puni - rak - spa - wa

speak-refl-bridge-prog-pluv-comp-nom-plun- we -just-always-also- vrb-3rdp+des-aff

‘I know it is desirable and necessary that all of us, including you also, keep communicating’

The piece-by-piece or ‘interlinear’ translations customarily provided in fieldwork linguistics (cf. VI.73) combine ordinary language with specialized functional labels (like ‘animate plural’) to suggest what the various morphemes contribute. In the Aymara data [15], the morphemes convey not just the content such as ‘speak’, but a rich palette of grammatical functions and aspects, such as reflexive (refl) (i.e. ‘speak to each other’), progressive (prog), plural of numerousness for the verb (pluv) and for the noun (plun), completive (comp) indicating sententious certainty, a marker for the third person of the desiderative aspect (3rdp+des), a transition from the opening verb to a nominal (nom) and then back to a verbal (vrb), plus a final affirmation marker (aff) that turns the whole cluster into a sentence type. The ‘idiomatic’ or ‘free’ translation shows English using ‘clausal framing’ to convey aspects such as sententious certainty and desiderative (‘I know’, ‘it is desirable and necessary’).

32. How has morphology fared on our three test scales proposed in II.28? Coverage faces the tough practical problem that the stock of minimal forms of a language could include the entire vocabulary of indivisible words plus word-parts. The preferred solution has been a theoretical division between the grammatical morphemes in stable, compact classes, e.g., the set of all verb inflections for ‘present’, ‘past’, and so on, versus the lexical morphemes in unstable, open classes, e.g., the set of all verbs or verb stems. Morphology would seek to provide full coverage of the grammatical ones as the proper ‘morphemes’, whereas the lexical ones would be lexemes and would be consigned to lexicology, a domain often left at the borders of linguistics (cf. II.38, 64). The ensuing division between ‘grammar’ versus ‘lexicon’ became an accredited dichotomy in °formal linguistics° (II.49), whereas °systemic functional linguistics° has projected the fundamental unity of the lexicogrammar apportioning the work of expression to grammatical or lexical resources in a characteristic way for each language or language type (II.63).

33. Morphology attained its best convergence through the fieldwork described in II.31, where theory and practice are always connected, and where all available clues must be exploited to discover the meaningful parts in transcriptions of actually observed utterances. This job requires acute skills in listening and transcribing, but also in relating utterances or utterance-parts to °cognitive and social constraints°, whether or not these might be described in ‘linguistic’ terms (II.27). Moreover, you must meet the practical challenge of moving from being a total ‘outsider’ for the community of speakers over to being an ‘insider’ who can speak the language well enough to interact in communicative practices (cf. VIII.9). Along the way, you continually test and refine your description, and adjust your theory accordingly.

34. Consensus is supported by the emphasis on identifying and isolating the morphemes in the data, which can extract some ‘cheap explanatory work’ from the linear distributions being visually displayed by a transcription into a reliable ‘phonetic alphabet’ (cf. II.23, 29, 38). One straightforward method is to repeatedly segment the data until no further meaningful subdivisions appear feasible, and then to classify the segments and the patterns these constitute. This ‘immediate constituent analysis’, if applied ‘in all observation of word-structure’, was expected to eliminate any ‘inconsistency of procedure’. You can, for instance, readily distinguish the bound morphemes that can occur only as a part of a longer unit from the free morphemes that can occur by themselves.

35. But even so, morphology is less congenial than phonology for the °classical program of mainstream linguistics° because the data are not nearly so tidy and compact. For example, Modern English has fairly few indisputable bound morphemes, such as the endings for plurals of nouns and for tenses and persons of verbs, or the endings such as ‘-able’ and ‘-ish’ for deriving words; in return, much of the ‘grammatical work’ is now done with the so-called Function Words (e.g., Articles, prepositions, and conjunctions) which, like bound morphemes in other languages, belong to small or closed sets and have key phrase positions but sparse or indeterminate meanings. The ‘lexical work’ is mainly done with the so-called Content Words (e.g., nouns, verbs, and modifiers) which belong to large or open sets and have flexible phrase positions but richer and more determinate meanings. A Content Word is more self-sufficient and more likely to be uttered alone (e.g., ‘fire!’, ‘run!’, ‘terrible!’) than is a Function Word (e.g., ‘?the!’, ‘?at!’, ‘?unless!’), except maybe to emphasize a contrast. Also, the Content Words are more likely to take on the remaining bound morphemes (the ‘inflections’), e.g., to signal a ‘declension’ of nouns with singular versus plural or a ‘conjugation’ of verbs with present versus past (IV.20, 47, 62, 82).

36. English also illustrates the problems for describing a language that has assiduously borrowed word-stems, prefixes, and so on, e.g., from French, Latin, or Greek, which are no longer recognized as meaningful units by many contemporary monolingual speakers. Should a morphological description include not just the more obvious units like the prefixes ‘in-’ and ‘im-’ for negation alongside ‘un-’, ‘non-’, or ‘a-’ but also the erudite units like ‘-pter’ (‘wing’) in ‘helicopter’, where speakers would more likely identify the final ‘-er’ as an agentive suffix (compared, say, to ‘propeller’) on the verb ‘helicopt’, which has in fact appeared in a recent English dictionary with the irritable label ‘back formation by false analysis’? Such etymological data would bring in language history and thus undercut the °mainstream program to describe language in a single stage of its evolution° (II.24(d)).

37. Still, these problems for theory have not kept morphology from making huge practical contributions by describing hundreds of previously undocumented languages through active fieldwork, such as that sponsored by the Summer Institute of Linguistics and carried out by Kenneth, Evelyn, and Eunice Pike, Robert Longacre, Joseph Grimes, and their many diligent colleagues and students (cf. IV.41; VI.73). This work is undoubtedly the most memorable and enduring achievement of modern linguistics. We have gained not merely a deepened and defamiliarized sense of the striking morphological diversity of languages, but also refined methods of investigation we can reapply to the more familiar, worked-over languages (cf. II.86, 108). We can also rework the classifications of languages begun by philologists like Rasmus Rask and continued by linguists like Edward Sapir, Joseph Greenberg, and Morris Swadesh (cf. II.19).

38. Looking to reapply the successful methods of morphology, the next domain in terms of unit size would logically be lexicology, whose theoretical units, the lexemes, correspond to the practical units of words; but this domain did not at all fit the linguists’ beloved conception of a uniform, stable, and abstract system. So the next domain taken up was actually the subsystem of syntax, which concerns the organization of phrases and clauses in ‘structures’ or ‘constituents’. Here, fresh problems cropped up. Consensus was hard to achieve about what the theoretical units, prospectively called syntagmemes (after the terms ‘phonemes’ and ‘morphemes’) or ‘phrase structures’, might be, and how they might correspond to the practical units ranging from just one word (e.g. ‘help!’) to a phrase and on up to an extended clause or sentence. Since the repertory obviously would not consist of minimal units, some new mode of theory would need to deal with complex units, Nor does it seem feasible to give an exhaustive, precise coverage of phrases and clauses; even the traditional division into ‘Subject’ and ‘Predicate’ can leave tricky residues, e.g., signals of the speaker’s viewpoint like ‘frankly’ (cf. IV.28, 206). Once again, the visual appearance of data written down does some ‘cheap explanatory work’ in indicating some divisions between words and between phrases (cf. II.23, 29, 34). But the linear sequence is not ‘rich’ enough: we can inspect the positions of items but not the relations among them — we can see where things are but not why they are there or where else they could be.

39. Evidently, the methods for segmenting data into formal units and classifying these into repertories supported °convergence and consensus° much better for phonology and morphology than for syntax. Linguists began to cast about for other theories and methods, particularly ones that would still treat language as a °uniform, stable, and abstract system°. So the system was redefined to consist not of a repertory of theoretical minimal units (II.30) but of a repertory of theoretical rules for arranging simple units into complex units. Here, the correspondence between theoretical units and practical units dramatically receded, insofar as these rules would not produce (or describe) the phrases and sentences themselves, but rather the ‘underlying structure’ of phrases and sentences. This new °paradigm° therefore felt authorized to reject the °data-driven and practice-driven methods of descriptive fieldwork° for ‘classifying and organizing data’ and ‘extracting patterns from a corpus of observed speech’, in favor of theory-driven methods for designing a °generative homework model of language by itself°. The model would be a generative grammar: a highly °technical construction° whose rules would °generate all the grammatical sentences of the language by assigning them structural descriptions°, and no ungrammatical sentences (cf. II.105). The grammar does not specify how people actually produce real sentences or why they say what they do (although discussions kept using terms like ‘producing sentences’); it only specifies the abstract structure of possible sentences. The homeworkers craftily declared it ‘unreasonable to demand of linguistic theory’ that it stipulate a ‘discovery procedure’ ‘for actually constructing the grammar, given a corpus of utterances’ (cf. VII.132, 326); ‘how one might have arrived at the grammar’ was ‘not relevant to the program of research’ — ‘one may arrive at a grammar by intuition, guess-work, all sorts of partial methodological hints, reliance on past experience, etc.’ Of course, practical discovery procedures were just what fieldworkers had always demanded of linguistic theory, and they were shocked to be called ‘unreasonable’ by people using ‘hints’ and ‘guess-work’.

40. The ratio between theory and practice in modern linguistics became more elusive than ever (cf. II.24). The °dichotomy in the classical program of mainstream linguistics between language by itself (‘langue’) versus language use (‘parole’)° (II.24) was redrawn between competence, i.e., what speakers of the language know, versus performance, i.e., what speakers actually do or say. ‘Competence’ belonged to an ‘ideal speaker-hearer in a completely homogeneous speech-community who knows its language perfectly’, and determines only what is grammatical, i.e., described by the grammar; ‘performance’ would determine what is acceptable, i.e., approved by real native speakers. This line of reasoning installed the idealized grammar as the formal model for competence (cf. VII.K.2). It was easy to invent plainly ‘grammatical’ or ‘well-formed’ examples like [16] and plainly ‘ungrammatical’ or ‘ill-formed’ ones like [16a], while still others seemed ‘grammatical’ and yet disturbingly odd like [16b] (cf. IV.129).

[16] Revolutionary new ideas appear infrequently.

[16a] New appear revolutionary infrequently ideas.

[16b] Colorless green ideas sleep furiously.

°Introspection and intuition° were expected to underwrite an explicit formal statement of the status of all such samples. But from that day to this, no such statement has been produced for any language. In hindsight, it seems evident that introspection and intuition cannot achieve in practice the task set down for it by the theory (cf. II.27, 65). No strong evidence suggests that the native speaker’s intuitive knowledge of the language is substantially more precise and uniform than his or her actual utterances and can be disconnected to make purely ‘grammatical’ judgements about invented sentences. Introspection is opportunistic, coming into practice when data are presented and evolving to adapt to viewpoint and focus, rather than subsisting as a stable theoretical system when it’s not being used.

41. At this stage in °mainstream linguistics° with °syntax as the model paradigm° (cf. II.30, 49), its status of a °normal science° fluctuated uneasily (II.28). The increase along the three °test scales° proposed in II.28 stagnated and gradually turned into a decrease. In theory, full coverage of language data was built right into definition of the °grammar: assigning a structural description to all the grammatical sentences° (II.39). But in practice, only modest sets of sentences could be so devised that their °grammaticality° was indisputable; the rest were sensitive to varying contexts and interpretations, which are presumably affected by °performance factors°. The homework linguists had once proclaimed that ‘descriptivist’ °data-driven practices° only ‘precluded the development of a theory’ (a ‘theory’ being of course a formalist grammar); the lesson now emerged, with high poetic justice, that ‘generativist’ °theory-driven procedures° preclude the description of adequate data in practice. The same grammar purporting in theory to provide total coverage of all the ‘well-formed sentences of English’ might in practice provide almost no coverage of a corpus of observed speech in everyday conversations, which would often look ‘ill-formed’ and which was always produced by real speakers who are neither ‘ideal’ nor ‘homogeneous’ and certainly don’t ‘know the language perfectly’ (cf. II.27, 54, 57, 66, 95). A paradoxical scenario arose: reality being claimed for a °technical construction° while rejecting the reality of human discourse — inclusive theory with exclusive practice (cf. I.6; II.27). And instead of becoming a genuine normal science, generativist linguistics could only be a normalizing quasi-science that cannot describe authentic data but only the data it has expressly normalized, thereby rendering them empirically undecidable (cf. III.172).

42. Convergence among the descriptions of data was impeded by the burgeoning theoretical apparatus of formal rules and notations and innovative terms. True to the ‘grammar’s’ name, each fresh rule-set ‘transformed’ the data’s appearance, sometimes closer to the ‘surface structure’ of sentences as we find them written down in ordinary orthography and sometimes venturing at varying distances into the ‘deep structure’ that was held to ‘underlie sentences’ and to ‘generate’ them through procedures of ‘transformation’, e.g., to convert Active into Passive. The prerogative to ‘transform’ the ‘surface data’ often led to divergent results, especially if, as was freely conceded, ‘surface structure’ is ‘unrevealing as to underlying deep structure’, and if ‘the grammar does not, in itself, provide any sensible procedure for finding a deep structure of a given sentence’.

43. Consensus was concentrated in the early phase when ‘generative grammar’ still had a ‘standard’ model’. Soon, the leeway for inventing theoretical rule-sets spurred proposals for competing models. The fuzzy connection between theory and practice in such models prevented using a °corpus of authentic data° to establish a consensus about the best model; instead, models were hotly debated on theoretical grounds (cf. VIII.22). Over thirty °technical constructions° still contend within the field, bearing trademark names like ‘case grammar’, ‘Montague grammar’, ‘lexical-functional grammar’, ‘X-bar theory’, or ‘government and binding’, and advocated on technical and formal criteria of design (II.27), which again do not favor consensus.

44. Perhaps the evolution of modern linguistics outlined in this section could be retold as the search for constraints, with constraint being broadly defined as any factor making some items or patterns of a language more or less likely than others. The Saussurian conception of ‘a system wherein everything holds everything else in place’ (‘un système où tout se tient rigoureusement’) suggests that a °language by itself° (‘langue’) consists of a complete set of standing constraints. The °generative° conception further assumes that the range of those standing constraints is formally circumscribed within the borders of the ‘grammar’. An alternative conception, sponsored by ‘systemic functional linguistics’ and related methods, sees language as a system (or ‘multi-system’) of evolving interactions between standing versus emergent constraints, and among linguistic, cognitive, and social constraints. Language is interposed like a layer inside a cake (Fig. II.3): society uses language to understand and

appropriate theworld , while the world passes via language into a socially shared ‘world-model’. The °classical program of °mainstream linguistics° would disconnect language from this configuration, as if rolling the layers apart (Fig. II.3, right side). °Language by itself° can be described in ‘purely linguistic’ terms only if it can hold firm and continue to subsist and operate upon its own internal, standing constraints (cf. II.54).

45. But can language by itself hold firm? It apparently can in some °sparse sectors° whose constraints are general and uniform and whose organization seems fairly °frozen°. In phonology, every °phoneme° is held precisely and uniquely in place by the physical constraints of articulation and by the mental constraints of being able to differentiate between elements that also differ in meaning (II.29). The physical constraints are the clearest, but the mental ones can be easily met by finding at least one contrastive data pair like ‘hid’ and ‘hit’, whose members do not mean the same thing; we needn’t state what they mean or how their meanings differ, we only need a ready consensus that the meanings do differ (cf. II.48).

46. Internal standing constraints also suffice for some fairly sparse and frozen sectors of morphology. For example, the ‘inflections’ of nouns and verbs can be

covered and classified without having to state their often complex relations to cognitive constraints (e.g., what might be the Agent of an Action) and social constraints (e.g., how Pro-Nouns can signal power or solidarity). Also, the distributions of morphemes can be firmly grasped by observing them in a °corpus of authentic data°, where the speakers themselves were respecting cognitive and social constraints (cf. II.33, 58, 75).

47. In syntax, however, where the constraints are more specific and diverse, the search was less successful. °Descriptive syntax° could not progress very far beyond morphology, e.g., in ‘phrase structure grammars’, and identified the linear positions of segments much better than the constraints that would put them there. Syntacticians kept looking ‘deeper’, whence the shift from fieldwork over to homework methods (II.39ff). Linguists apparently assumed that purely syntactic constraints corresponding to ‘rules’ would emerge most clearly from invented data out of context, since authentic data in rich contexts obviously obey non-syntactic constraints too. Stating which sequences (or ‘sentences’) can or cannot occur in theory should prove easier than stating which ones actually did in practice. Predictably, the ambition to make syntax stand on its own constraints highlighted the frozen islands of language, i.e., the stabilized formal patterns of °sparse standing constraints° such as ‘article + noun’ in English. Yet even in languages which, like English, have a fair quantity, these frozen islands by no means suffice for continuing increases in °coverage, convergence, and consensus°. Without realizing it, generative syntax had taken on the unworkable task of freezing the whole system, and each attempt necessarily got partial and divergent results (IV.3). Even simple invented data (e.g., ‘John is eager to please’, II.66) may refuse to freeze over: the more you fixate and analyze them, the likelier you are to notice other aspects or interpretations.

48. This predicament encouraged widening the search for constraints to encompass semantics, the investigation of the meanings of language. Early theorizing had implicitly disconnected semantics from the rest of linguistics by insisting that the relation between form and meaning is arbitrary, i.e., not motivated by any natural bond. Intriguingly, this move deviated from the °mainstream program° by appealing to the history and evolution of language: how the meanings came to correspond to the sound patterns we now observe. But the relation between form and meaning is not all arbitrary within the system and the consensus among speakers (II.62); it eludes linguistics because it is always evolving in multiple dimensions on several levels. In phonology, the phonemes need merely differentiate meanings (II.29, 45). In morphology, the morphemes grammaticalize meanings, but the meaning may not be decided until a morpheme gets used, e.g., the English Possessive ‘‑’s’ that might indicate many relations beside possession, e.g., part (‘the car’s engine’), temporality (‘tonight’s feature film’), locality (‘Hollywood’s fashion industry’), and so on. In lexicology, the lexemes lexicalize meanings that are typically stable though often adaptive, e.g. ‘freedom’ (VII.10, 18, 31). In syntax, the syntagmemes linearize meanings that are also grammaticalized and lexicalized. This scheme, which will be enriched later on (II.62; III.92, 203, 232ff), does not foresee a level or domain for ‘meanings by themselves’, fully separated from these other levels: such a level might indeed look quite ‘arbitrary’.

49. At all events, semantics long led a shadowy life within linguistics, and was often left to the philosophers. Some linguists even vowed that meaning should not be a part of their ‘science’ at all. Others undertook to develop a semantics closely modeled upon the better-known levels. When phonology was the °model paradigm° (II.30), ‘structural semantics’ postulated a repertory of sparse-looking theoretical minimal units called ‘sememes’ (or ‘semes’ or ‘semantemes’), such as ‘± Animate’ or ‘± Human’. When syntax became the model paradigm (II.41), this scheme was taken over into ‘semantic features’ (or ‘markers’) specified by formal rules and situated in the ‘lexicon’ rather than in the ‘grammar’ — these two being considered separate levels or components, as they already had been in morphology (cf. II.32, 75; IV.14). Neither approach fared well on the °test scales of coverage, convergence, and consensus°. Research centered on programmatic demonstrations with meagre examples, and even simple analyses (e.g. ‘kill = cause + die’) raised thorny disputes. In retrospect, this outcome could have been predicted. Of all the aspects of language, meaning is the hardest to ‘freeze’ and suffers the most through °disconnecting language from people’s knowledge of world and society° (II.74). We are left with no way to tell what practical units these theoretical units might correspond to: ideas, human actions, real objects and events, cultural artefacts, and so on. A ‘sememe’ or ‘feature’ like ‘+Human’ cannot be a stable, determinate unit: how complex or simple and how abstract or concrete it is depends on the context, e.g., on whether it occurs in the discourse of biology, anthropology, philosophy, religion, politics, law, or medicine (cf. III.93). Trying to analyze meanings out of context or else for all contexts are two equally impossible jobs — the one having too few constraints and the other too many — and linguists get vague or conflicting results (cf. II.53f; III.32). Meanings can converge and linguists can reach a secure consensus only by assembling a representative set of discourse contexts, where the constraints have been applied by real speakers (II.65, 75).

50. Nor has it been easy to reach a consensus about how semantics should interact with syntax. After withdrawing his early claim that ‘grammar’ is ‘autonomous and independent of meaning’, Chomsky’s ‘standard model’ gave the ‘semantic component’ the job of supplying formal ‘rules’ for ‘interpreting’ and ‘disambiguating’ syntactic sequences that had already been ‘generated’ (‘structurally described’, II.39) by the syntax. Because this arrangement blocked the semantic constraints from helping to construct the sequence, a counter-proposal was soon made to give the ‘generative’ role to the semantics, and to postulate the ‘logical form’ for the ‘deep structure’ of the sentence before the syntax kicks in. But consensus also broke down about how to represent a ‘logical form’: logic pursues sparse issues like ‘truth’ and ‘validity’ rather than the rich ordinary meanings of words or utterances (cf. II.10ff). The ‘logical forms’ shown in sample analyses glibly retained undefined ordinary words, e.g., in this ‘semantic representation’ for a sentence from James McCawley:

The formula states that it is true for all members (x’s) — ‘"’ being a ‘universal quantifier’ — that each x loves x’s wife. The formal treatment covers only the assignments into sets, an issue logic can handle (II.11). The rich meaning of the crucial ‘love’ is not semantically or logically analyzed into units, and attempts to do so would hardly converge.

51. Further constraints were plainly needed, and the stage was set to welcome pragmatics, which had been defined by linguistics and semiotics as the study of the uses of language in a three-part scheme alongside ‘syntax’ and ‘semantics’. Fended off by the °staid division between language by itself versus language use°, pragmatics had made its academic home mainly in philosophy. Linguistics now admitted pragmatics on the condition that, like semantics, it would respect the ‘formality’ prescribed for syntax. So pragmatics was to supply further sparse constraints in the guise of ‘rules’ for ‘interpreting’ and ‘disambiguating’ sentences by specifying what a sentence ‘implies’, ‘entails’, or ‘presupposes’, and how its meaning (sometimes!) depends on the speaker’s intention and the hearer’s acceptance. What appears to be a simple statement about reality, e.g., Rodolfo’s ‘c’è freddo fuori’ (‘it’s cold outside’) in Puccini’s La Bohème, can be intended and accepted to be a proposal to stay inside; Mimi’s answer ‘ti starò vicina’ (‘I’ll be near you’) reacts by implying that she’ll keep him from feeling the cold outside.

52. A catchy motto for pragmatic inquiries was ‘how to do things with words’ (John Austin), and not just how words are formed and arranged or what words mean. The notion of speech act was adopted from ordinary language philosophy, a term contrasting with the more technical or formal language philosophy of logical and mathematical systems (cf. III.167). Here, research sought to specify the conditions under which an utterance act or a locutionary act corresponds to a propositional act of stating a content or message, an illocutionary act of performing a discourse action (e.g. promising, threatening), and a perlocutionary act of eliciting an effect on the hearer (e.g. getting compliance). Whereas the propositional act foregrounds the constative function of making statements about the world, the illocutionary act is most prominent in the performative function where saying and doing fully converge (e.g. ‘pronouncing this meeting adjourned’).