Printer Friendly

What every grammarian knows?


The grammarians of antiquity, unlike some of their modern counterparts, seem to have had little interest in investigating `what every speaker knows', at least as a large scale project, consciously articulated and embarked on. The object of such a project would be to determine what constitutes such knowledge--or mastery, or cognition, or whatever name it is given--in actual speakers. An alternative goal would be an account of something knowledge of which would count as knowledge of the language in question. even if this is not what actual speakers know. No ancient author seems to have set himself either task; indeed, the whole question of what it is to know a language seems to have been raised only very occasionally in classical antiquity. (Protagoras' allusion to a `teacher of Greek' in Plato's dialogue (327e2-328al) is probably the best known of these rare discussions.) Kaimio is surely justified in making the still more sweeping observation that `The concept of language knowledge was rather vague amongst ancient peoples' (1979: 316).

I propose to investigate one aspect of this remarkable disparity between the indifference of antiquity and the zeal of today. It is only comparatively recently that grammarians have been able, if they so wish, to see themselves as attempting to reconstruct the internalized grammar which is a speaker's knowledge of his or her language, and, ultimately, the innate and arbitrary `universal grammar' of the human race; or as seeking to fix rules knowledge of which would count as knowledge of this or that language, whatever the relation between these rules and the ones speakers actually know. For such a project, speakers' intuitive linguistic behaviour, including their own judgements about the cardinal properties of utterances--grammaticality, ambiguity, synonymy, and so on constitutes the primary data. Ancient grammarians, in contrast. decidedly had no such conception of their subject-matter or their goals: and some modern practitioners of and commentators on linguistics-would no doubt compliment them on their good fortune in this regard.

For it can hardly be claimed that the nature, scope, structure, and origins of ordinary language knowledge are now so thoroughly understood that we are in a position to evince justified indifference, or contempt, toward the efforts of earlier generations: the problems which critics such as Baker and Hacker (1984: 242ff.) have detected in `realist' grammars are certainly grave. How (above all) could a speaker be said in any sense to `know' or have `mastered' rules which, when formulated in the abstruse argot of the linguist, are unintelligible to all but the professional, and are, ex hypothesi, inaccessible to the user? In what language are these rules framed, and what are its rules? Is the very notion of a rule which the originator of the supposedly rule-governed activity does not and cannot (even in principle) cite to justify and explain his or her activity even coherent?

The grammarians of antiquity, when they set about constructing rules for correct discourse, had no such anxieties. Indeed, speakers' intuitions were, typically, not even regarded as setting extensional limits to correctness: words or sentences correct in usage could be judged incorrect by another, preferred standard; and vice versa. Yet we shall see that by their very failure to recognize the abilities of ordinary speakers, participants in the ancient grammatical enterprise cut away the ground from beneath their own feet much as today's realist grammarians may have done, in their case by failure to recognize the nature of rules and rule-governed activities. Ancient grammarians, in contrast, had very clear ideas about linguistic rules, yet seem never even to have considered the possibility that ordinary speakers had access to them.


The social factors which gave rise to grammar in the ancient west, and brought it to prominence as an integral, if comparatively humble, part of the education of the elite, in themselves tended to turn grammarians' attention away from ordinary usage, especially of the low-born and uneducated. (Kaster has well emphasized the professional's feelings of superiority over the vulgar, cattle-like masses (1988: 17).) Ordinary usage was itself typically regarded as sub-standard, as shot through with irregularities and inconsistencies. The primary focus of grammarians' efforts and expertise was the literary canon, its textual correction, exposition, and assessment, at all levels of sophistication and originality. Thus, for example, the literary text is central to all four parts of' `old-style' grammar as described by one of the Dionysian scholiasts (Grammatici Graeci I iii, p. 12, 11. 3ff.); Varro's fourfold division of grammar betrays a similar bias (Diomedes, Grammatici Latini [G.L.] I, p. 426, Keil); and another scholiast claims that the reason for the existence of grammar is the `lack of clarity of poems and prose works' (D. T. Sch., 113.15-20; and cf. e.g. Robins, 1957: 70). Admittedly, `ordinary usage' would commonly appear alongside analogy, etymology, the various dialects, antiquity. and established literary language as a criterion of linguistic purity ([GREEK TEXT NOT REPRODUCIBLE IN ASCII]), latinitas): but, as the jargon itself --[GREEK TEXT NOT REPRODUCIBLE IN ASCII]; its Latin calque consuetudo reveals, that criterion had reference to linguistic practices, not to their cognitive basis. (The terms [GREEK TEXT NOT REPRODUCIBLE IN ASCII] and [GREEK TEXT NOT REPRODUCIBLE IN ASCII] are also used of literary `usage', often that of a particular author (as at Sextus, Against the Professors [M. I 205; at 203 `the most approved and oldest usage' from which analogy is to be learned is that of Homer, so that the two uses merge); see further Siebenborn, 1976: 86, 90. Except where indicated, these terms should be taken here to refer to `ordinary usage'.)

One recurring question, in antiquity as in modern scholarship, has been how grammarians regarded, and supported their claims to, professional knowledge or competence. Sextus Empiricus' famous critique of grammar (M. I 41ff.), for example, contains a lengthy report ([sub sections]57-89) of a range of positions adopted, not only on grammar's scope and methods, but also on its status as a discipline. The appearance of this catalogue here is in itself unsurprising. given Sextus' avowed, Skeptical, intention, of arguing against the existence of a genuine grammatical `expertise', [GREEK TEXT NOT REPRODUCIBLE IN ASCII] ([sub sections]89), as of all the other standard components of the usual elite education ([sub sections]7f). Other sources, however, readily confirm continuing anxiety about the epistemological basis and respectability of grammar, Dionysius Thrax' own definition of grammar as a sort of `empirically-based knowledge'. [GREEK TEXT NOT REPRODUCIBLE IN ASCII], having apparently sparked off the debate. at least as we know it (D. T. Sch. 14.24ff., 165.16ff., 166.25ff., 300.15ff.; Sextus. M. I. 60); and Crates also seems to have put his oar in, by making true grammar. [GREEK TEXT NOT REPRODUCIBLE IN ASCII], dependent on the science of language (M. I, 79).

The origins and early history of grammar, particularly of that variety of it which the ancients called `technical', and the precise contexts within which such disputes about professional and epistemological status occurred, are still very much subject for debate, raising questions far too large and complex to be more than touched on here. (The main problem, of course, is the dearth of early, reliable evidence (as already noted by Baratin and Desbordes, 1987: 42; Frede, 1987b: 339; Kaster, 1988: 137), in marked contrast to the wealth of secondary literature.) 1 shall concern myself with such difficulties only insofar as they bear on ancient views about language knowledge. But two things at least stand out clearly.

First, grammar took an unconscionable time to distinguish itself from the already established and powerful disciplines--philosophy, rhetoric, and music above all (e.g. Sextus, M. I, 80)--with which it had enduring links, and never in fact severed those links entirely. Grammarians thus had, and for a considerable period, a vested interest in demarcating their field of activities, and in defining the conduct of them. Yet they seem never to have felt the need to distinguish what they knew, or the manner in which they knew it, from what ordinary users know: this last provided only ammunition for critics of grammar (as in Sextus, M. I, 66), not the grounds for such discriminations and definitions in the first place.

Second, grammarians giving instruction in Greek or Latin taught, not (a) language per se, but a certain sort of language, one endowed with certain covetable qualities--purity imprimis. Such grammarians as Apollonius Dyscolus and Priscian, who made Greek or Latin the object of a more theoretical, abstract, and sophisticated study, were no less concerned with defining and explaining what was linguistically pure or correct. There may, then, have been conceded to ordinary speakers some special area of linguistic activity, but not one of which there could be genuine `knowledge', as knowledge was understood. On the other hand, speakers may have been dismissed as not genuinely knowledgeable, on the grounds that their grasp of the pure language, the language with which the grammarians concerned themselves, was insufficiently sure, precise, comprehensive, or articulate to count as knowledge proper--unless and until they had acquired the very knowledge which grammarians claimed to impart (cf. Kaster, 1988: 205).

The claim that the ancient grammarians' lack of interest in ordinary language knowledge needs investigation cannot, however, be rejected merely because no ordinary user could have counted as an expert on literary texts--the texts which, as we have seen, occupied so crucial and central a place in the grammatical enterprise. It goes without saying that the expertise, or rather nexus of expertises, required to establish a text, explain its unusual words or phrases, point out its tropes and figures, elucidate its allusions to history, mythology, customs both secular and religious, and so on, and offer judgement on its merits and defects, could never have been part of the untutored speaker's knowledge of Greek or Latin could never have been a necessary condition for a person's being a native speaker of either (or any) language. In this discipline, ordinary people's intuitions about their usages would have been quite irrelevant, ordinary usages themselves largely so, although good or approved usage, or the characteristics of a particular dialect, might have been invoked in order to defend or criticize this or that reading of a piece of literature. No difficulty is raised for us because grammar was not restricted to relatively esoteric. or `expert', textual and literary criticism and exegesis.

I have already conceded that this is not the place to enter into the complex and uncertain history of grammar in antiquity. It is enough to observe that some portion of grammar came to be devoted to the setting up and application of criteria of purity, especially, if not exclusively, as one of the major `excellences of language',[GREEK TEXT NOT REPRODUCIBLE IN ASCII], alongside clarity, appropriateness, and elegance, or some variation, or subset. hereof. (This variant of ancient stylistics is now well-enough known: see. for example, Siebenborn, 1976: 35ff.; Frede, 1987a: 309-12: Baratin and Desbordes. 1987:43-6;Atherton,1988:419f.)Whatever its history and antecedents, the so-called `technical' grammar, which took linguistic correctness as one of its principal concerns. came to be recognized as an essential tool of grammar as a whole (e.g. Varro, fr. 109 G.-S., D.T. Sch. 115.8, 164.9, 170.17, and cf Barwick 1922: 224f.) or as a subdivision of it (as described by Sextus, M. I 91f.). Its specialized competence seems already to be present in the Dionysian conception of grammar in the guise of working out analogy', [GREEK TEXT NOT REPRODUCIBLE IN ASCII] (Grammatici Graeci I i ars gr. 5.4-6.3; cf. esp. D.T.Sch. 169.24f, with Robins 1957: 83, 1976: 334; Lallot 1989: 80; Swiggers and Wouters, 1994: 2 n. 4). In the 1st century B.C.E., Asclepiades of Myrlea apparently identified [GREEK TEXT NOT REPRODUCIBLE IN ASCII] as one of the three parts of grammar (Sextus, M. I 252, with [sub sections] 74, and cf.[sub sections][sub sections]248f., on Tauriscus, a pupil of Crates'). The Stoics, too, make implicit reference to a [GREEK TEXT NOT REPRODUCIBLE IN ASCII] of Hellenism (Diogenes Laertius VII 59, with Frede, 1987a: 310).

Technical grammar had a natural interest in the ancient counterparts of the subjects today termed phonetics, phonology, morphology, syntax, and semantics. Although technical grammar's relations with other grammatical activities remain unclear, it is obvious that criteria of purity were a central preoccupation of many grammarians, whether in teaching, in editing and expounding texts, or in constructing grammatical accounts of Greek or Latin: see, for instance, Sextus, M. I 91f., 97, 175, Quintilian, inst. vat. I iv 2, v 1, ix 1, and also Varro, de lingua latina (L.L.) VII 6, with its distinction between historia and ars. Such texts also make clear that this type of grammar was regarded as `technical' in quite a strict sense, that is, as constituting a [GREEK TEXT NOT REPRODUCIBLE IN ASCII] or expertise: a definite body of knowledge, about a specific subject-matter, systematically articulated and organized, publicly accessible to all having the appropriate training, and directed to a practically useful end achievable by the application of fixed and defensible procedures. Of course, grammarians, like other professionals in antiquity, continued to argue about what expertise is, what types of expertise there are, and how they differ from all other forms of cognition, such as knowledge by experience([GREEK TEXT NOT REPRODUCIBLE IN ASCII]) or scientific knowledge ([GREEK TEXT NOT REPRODUCIBLE IN ASCII]). But the outline I have given here would, I think, have been widely acceptable, and it will be adequate for the moment.

The importance of the concept of linguistic purity to the distinctness and distinctiveness of technical grammar as a (sub-)discipline is hardly to be exaggerated. Purity was almost universally recognized as an essential property of good language. Exceptions to this rule are rare and distinctly unorthodox: I know only of Chrysippus' recommendation to overlook solecisms in a speech under certain circumstances (Plutarch, st. rep. 1047B, with Atherton, 1988: 420, 422f.), which Plutarch finds suitably shocking, and Galen's argument that clarity is the sole linguistic excellence (On linguistic sophisms [ea. Ebbesen, 1981: vol. 2: pp. 1ff.], chs. 1 -3, with Atherton, 1993: 201-4). The standard view was that without purity all other qualities of style--typically lucidity, appropriateness, and ornament--would be pointless (e.g. Cicero, de oratore III 38; D. H. Pomp. 3). A rhetorician might well regard pure speech as a mere sine qua non, not a positive merit in its own right (cf. Cicero, Brutus 133, 140, de or. III 38), but rather one to be mastered before the real business of rhetorical instruction began (ad Her. IV 17; Quintilian I ix 6, VIII ii 1f.). Any number of earlier speakers had, indeed, acquired pure speech effortlessly at their mother's knee (Cicero, Brutus 258f., cf. 210f.): later on, by contrast, linguistic purity has to be deliberately fostered, by careful provision of a home environment in which only pure language is to be heard, by wide reading of the classics, and by appropriate schooling (Cicero, Brutus 252, de or. III 48; Quintilian I i 4-7, 12-14, 37, iv 6-12, v 17-33).

Practice need not have followed precept, here or anywhere. But our concern will be with the theory and methodology of technical grammar, and with its image and status among,t grammarians and other professionals. In their different ways both the local grammarian, struggling to instruct schoolboys in the names of the cases or the correct use of the article, and the [GREEK TEXT NOT REPRODUCIBLE IN ASCII] par excellence, Apollonius Dyscolus, attempting to catalogue and explain the properties, behaviour, and syntactic relations of each of the parts of speech, made linguistic purity a pillar of grammar (cf. Bonner, 1977: 1981T.. 204ff.); and hence, too, a pillar both of elite education, and of the study of language.

Our first concern will be to expound and understand the standard criticism of ordinary language as irregular and eccentric. Next, we shall see how, under the influence of the epistemological professionals, the philosophers, and their preoccupations, the conceptions of knowledge which technical grammarians tended to adopt would work against the inclusion of ordinary speakers in the select band of genuine knowers. Such epistemological prejudices ensured that ordinary users' knowledge of their native tongue--widely regarded today as one of the wonders of the world--could never be accorded the high theoretical and heuristic status it is granted now even outside `realist' grammars.

I shall not be asking the absurd, and uninteresting, question, why ancient technical grammarians failed to anticipate the now widely influential, `realist' model of grammar sketched earlier--as if that were the obvious, the only acceptable, form of grammatical activity. My primary concern is with another problem: why the technical grammarians tended not to see ordinary users--those who, as Quintilian has it (XII x 40), use `everyday speech [`cotidiano sermone'] ... with friends, wives, children and slaves'--as in any way authorities or experts on language despite their apparent proficiency in the actual business of talking and listening, except, trivially, as `experts' on what was or was not current usage. All grammarians unthinkingly accorded themselves `expert' status in this looser sense, whenever they affirmed that such-and such a usage exists; but they did so without compromising--indeed, while confirming--their claims to specialist, expert, status as grammarians (cf. Kaster, 1988: 178). (We can, therefore, draw the sting of one potential objection to the inquiry l propose: that, as Versteegh [1987] has argued, ancient grammarians had no truck with the usage of the great unwashed--with the wholly, or almost wholly, unschooled and illiterate majority. For the crucial distinction for us is that between lay and professional linguistic behaviour and knowledge, not that between high, middling, and low dialects of Greek or Latin. Of course, if Versteegh is right about the nature of the grammarians' `ordinary usage' and [GREEK TEXT NOT REPRODUCIBLE IN ASCII], the circularity of appeals to this (educated) usage as a criterion of purity would be hard to deny: but that is someone else's problem.) By investigation ancient neglect of ordinary, non-professional, linguistic cognition, we may, however, begin to see why the modern realist model of grammar could never have been a possible model for ancient theorists of language.


The adoption of usage as a criterion of purity seems to have been both early and extremely widespread. Grammarians began to claim authority over all linguistic phenomena, including ordinary speech (Sextus, M. I 76 [shares], 79 [Crates], 84 [Demetrius Chlorus]; cf. 64), and there at least it would be most odd if usage lacked authority. Concessions to some sort of usage are already implicit in the Stoic definition of Hellenism, with its allusion to the [GREEK TEXT NOT REPRODUCIBLE IN ASCII] (D.L. VII 59; cf. Frede. 1987b: 310-11). and in Aristarchus' urging not necessarily explicit--that forma] analogy cannot ov-erride usage (Varro, L.L. IX 1. with Siebenborn, 1976: 30f; Frede. 1987b: 340-42).

On the other hand. adopting usage tout court as the sole criterion seems scarcely to have been an option. Only the Stoic-influenced Crates is supposed to have championed anomaly and usage (supposing it to be common usage, not literary idiolect. which he had in mind). Proper discussion of the evidence about Crates' views on language would need a book to itself, and I shall mention here only the fact that the sort of anomaly of interest to Crates seems to have been in part the semantic variety (if Varro is to be believed, L.L. IX 1); so that, as the Stoics seem to have been unpersuaded that language needed to be restored into its pristine state, before anomaly, whether semantic or formal, took hold (cf. Atherton, 1993: 93f.), Crates' reported defence of anomaly and usage may have amounted to no more than a refusal to play the revisionist game. Or his point may have been that semantic properties cannot be appealed to when constructing or applying rules for formal analogy (examples of this approach to analogy at Quintilian I vi 12f.). In any case, we shall be focussing on the more orthodox attitudes to linguistic correctness; and here the common attitude came to be, almost always, one of compromise, whereby a standard matrix of correctness criteria inevitably produced a relative linguistic purity: relative to author (literary idiolect), to location (or assumed location) (dialect), to period (vetustas), to literary context (auctoritas) and perhaps to social class as well. Only analogy and etymology will not prove relative in this way, being internal to a language and to its history, and their nature and effects can only be overlaid or distorted, not essentially changed or eradicated, by usage. As for usage, even when it is praised, it is consistently, and significantly, the better' sort (e.g. Quintilian I vi 43 - 5). It seems, in fact, to have been assumed quite early on in the history of grammar that ordinary usage at least appears to be formally (as well as semantically) irregular. The debate was rather over how its irregularity should be managed (cf. Siebenborn, 1976: 108ff., who argues for different approaches from Greek and Latin grammarians).

On the one side, then, Quintilian allows that, while pure Latin may be corrupted by barbarism and solecism, such defects may be overlooked because they appear in ordinary speech, or in the canon, or because they are old, or even because they closely resemble figures and tropes (I v 5, cf 40). (That tropes and figures are, so to say, acceptable poetic barbarisms and solecisms, became something of a commonplace: cf. Baratin end Desbordes, 1987:48-51,55-62;Kaster, 1988: 151f.;Atherton, 1993: 485.) Varro is reported (G.L. I 50, 489) to have espoused four criteria: natura (that is, etymological correctness of words), analogia, auctoritas, and consuetudo. (We shall be examining his views on usage a little later.) Another strategy was to explain irregularities away as not really irregular at all, as being in fact rule-governed themselves. The most impressive effort in this direction is Apollonius Dyscolus' attempt to present the received usage of the canon as conforming to the same rules as are dictated by reason (see esp. pron. 113.1 7ff., adv. 199.16f. Schneider, synt. I 60, p. 51.1ff. Uhlig, with Blank, 1982: 14; Siebenborn, 1976: 54f.).Yet here some of the phenomena could prove recalcitrant: Apollonius is forced to admit, for example, that the common use of plural nominatives with singular verbs is incongruous (synt. III 53, p. 315.16ff., esp. p. 319.1f Uhlig); and it is unclear whether the prospect of rehabilitation could ever be held out to ordinary, colloquial, usage.

Most radically of all, irregularities could be eliminated from usage by a programme of linguistic revision and improvement. That some theorists, and some speakers, genuinely wanted to do away with anomalistic usage is-suggested by, for example, Sextus, M. I 1 76ff. (cf. 97-9), and Quintilian's approving reiteration of what seems to be a hackneyed saving, aliud esse Latine, aliud grammatice loqui', `speaking Latin is one thing. speaking grammatically another' (I vi 27). Caesar is reported to `correct, by the application of reason, with the help of pure and uncorrupted usage, usage which is corrupt and impure' (`rationem adhibens consuetudinem vitiosam et corruptam pure et incorrupta consuetudine emendat', Cicero, Brutus 261), although he also warned his readers to avoid any word which is `unusual' (insolens) or strange' (inauditum') (fry. 2 Funaioli) Such `rational' linguistic revisionism is thus, perhaps. to be distinguished from the deliberate, blithe, anachronism of Sisenna, who reportedly wanted to be `corrector, so to say, of customary language', quasi emendator sermonis usitati', and, by his own lights, spoke Latin `correctly', `recte', precisely in speaking `inusitate', `unwontedly' (259-61).

In brief, the trend was to accept usage of (educated or elite) speakers as a check on unbridled regularity or analogy, but this was,-typically, because such usage (on the whole) chimed in with regularity: it was not an autonomous source of regularity, but an autonomous source of correctness, distinct from analogy and, in practice, making inroads into analogy's linguistic territory. The analogist position retailed or constructed by Varro has it that `we will follow those who call us to attend to usage, if it is the correct [`rectam'] sort: for in this too there is analogy [`in eo quoque enim est analogia `d; but if they summon as to attend to corrupt [`depravata'] usage, we will not, unless we have to, any more than we follow bad examples in other matters: for we do follow them, too, against our will, if we are under duress' (IX 18). The crucial point for us is that `better' when applied to usage tends to mean `(more) regular' (`in eo quoque enim est analogia'): and this is so whether the theorist bows to usage under pressure, or sees it as a perfectly acceptable guide to correct language. It is surely significant that it is usually the rhetoricians who remind us that usage is prior to, and gives rise to, analogy, for they have a vested interest in language which has the desired effect on, and is acceptable to, their audience. Quintilian, for example, points out that analogy, you see. was not sent down from heaven, to give a shape to language, when human beings were first created: rather, it was discovered after they started to speak, and it was noticed what was inflected in speech' (`Non enim, cum primum fingerentur homines, analogia demissa caelo formam loquendi dedit, sed inventa est postquam loquebantur, et notatum in sermone quid quomodo caderet', I vi 16). It is better to err with the eloquent than keep to the straight and narrow with the grammarian (I vi 2), and usage is `Language's surest teacher', `certissima loquendi magistra' (I vi 3). Yet Quintilian is cockahoop when he finds analogy, auctoritas, and vetustas in alliance against his critics (I vi 11), while readily admitting that usage can overcome the force of linguistic `law' (v 29).

It is, in part, acceptance of analogy as an independent linguistic force which allows us to understand the typical grammatical attitude to ordinary language. The remarkable theory of language presented by Varro in his de lingua latina provides an outstanding illustration of these standard attitudes precisely because of what-seems to be striking originality in other ways. Perhaps uniquely in antiquity, Varro recognized the mastery displayed by ordinary speakers, even slaves, of the inflectional patterns which make language learning possible and language use easier and more effective (VIII 3-6; Quintilian cites the ease with which children learn a first language as evidence that they will find learning anything easy, but inflection is not mentioned [I xii 9]). Mistakes here are rare, just as those who originally imposed names made only a few errors, such as giving plural names to single objects (L.L. VIII 7), and newly learned words are typically correctly inflected without hesitation', `sine dubitatione' ([sub sections]6). The populus or linguistic community should, properly, adhere to the natura or ratio of its language, but, if it does introduce usages at variance with analogy, individuals--with the exception of poets, who are allowed more licence than the rest of us, and can defy this dictatorship of the people--must follow the usage of the community none the less, and not the rules of analogy (Sisenna, and even Caesar, would certainly not have earned Varro's approval: IX 5f.).

Varro also recognizes the creativity of speakers who form new words from old (VIII 21). (Macrobius represents Servius as making up new words on archaic patterns (Sat. I iv 20-27), as Kaster [1988: 171f.] points out: but this conscious, literary, creativity, comparable rather to the well-established licence granted to poets to invent words, is quite different from untutored speakers' inventions.) What Varro calls declinatio voluntaria or `free word-formation', which we would call derivational morphology, in no way prohibits the action of declinatio naturalis or `natural word-formation', which we would call inflectional morphology (e.g. VIII 21f., X 15; on the two types of declinatio, see, for example, Taylor, 1974; Itkonen, 1991: 195-8). Indeed, it is their combination which allows Varro to explain how both analogy and anomaly play a part in language, the former being seen at work more in the words which speakers make up as required (say, when giving a slave just purchased a new name), the latter more in inflection (VIII 23, IX 2). Both types arise from usage (IX 2f), a point comparable to Quintilian's quasi-legal description of the matter: analogy `relies, therefore? not on reason, but on example, nor is it a law laid down for language, but careful attention to it [or: ` ... not a law for language, but a practice observed in it'], so that nothing created this very analogy but usage' (`Itaque non ratione nititur sed exemplo, nec lex est loquendi sed observatio, ut ipsam analogiam nulla res alla fecerit quam consuetudo', I vi 16; and cf. V ii 1 on exempla in law, a form of non-technical proof. with III vi 86, for the standard general distinction between rational and legal status).

Varro's explanation of the two sorts of derivatio, and, in particular, the autonomy of `natural word-formation', does raise a number of problems, most notably an apparent failure to allow for either diachronic or synchronic variation. But from our perspective there is an equally significant omission: the failure to pursue the observation that speakers have (almost perfectly) mastered and constantly put into practice their mastery of the system of declinatio naturalist As for word formation, Varro's gaze is fixed almost exclusively on the language, on its natura or organic systematicity, or on its ratio, its rational patterning, and he does not dwell on the fact that even the lowliest of ordinary speakers have managed to acquire not only a basic lexicon--to which that portion of grammar known as historia is devoted--but also the set of inflectional rules which the ars of grammar, its technical portion, aims to collect and collate (VIII 6).

For Varro, language in its true essence has an existence, an authority, and a creativity, independent even of the language community which, in some sense, `ought' to make use of it (IX 5f). It is noteworthy that language in this sense seems to be immune from change or variation: `the nature of verbs and nominals <is> unchangeable, and it hands on to us nothing more, nor less, than it received' (`nature verborum nominumque immutabilis <est> nec quicquam aut minus aut plus tradidit nobis quam qlLod accepit'), so that any error `is proved wrong, not by the power of analogy, but by the inner structure of the very nature <of language>' (`convincitur non analogiae virtute sed naturae ipsius constitutione') (fr. 115 G.S.). Of course, `it is one thing to say that words have analogies, another to say that one ought to use analogies' (`aliud est dicere <esse> verborum analogias, aliud uti oportere analogies, IX 4). But that is precisely to affirm that analogy exists in language regardless of the utterances people actually produce.

It is impossible today to judge just how innovative Varro's theory of language was--earlier and contemporary material simply does not exist in sufficient quantity for informative comparison with--it but his recognition of (part of) the content of speakers' knowledge may have been unique. At the same time, his (very brief) remarks on the topic are quite consistent with what we learn elsewhere about usage as a criterion: although it may be obscured by usage, the regularity of the set of patterns of inflection helps lend respectability to some portion of that usage; while usage. for its part' has in its favour not only intelligibility to audiences, or the authority conferred by its provenance--the discourse of the elite--but also this limited overlap with analogy. The grammarians' focus is always on the discrepancies between what is common parlance and what is regular, for it as here that usage comes into its own as a criterion of correctness. It will be worth pursuing a little further the role which analogy came to play for the technical grammarian.

No ancient grammarian saw ordinary speakers' utterances, including their intuitive judgements regarding their own and others' utterances, as linguists today typically see them: as the principal source of information about the language system. This methodological and heuristic principle will hoed good even if it is unclear how the rules constructed by grammarians can be the rules which speakers have internalized. My point is not merely that grammarians in antiquity did not take intuitive linguistic behaviour even as setting the boundary between what is correct and what is incorrect. We must also keep in view what it was that technical grammarians were seeking to formulate when they turned to analogy to construct rules for inflection. It would be gravely misleading to regard their ultimate goal as the description, even the piecemeal description, of a language `system', which is how modern linguists typically refer to the object of their attentions. There are several cardinal differences, but I shall mention only one here. The modern descriptive grammarian does not offer the system' as a mode] for future linguistic behaviour. The system is, admittedly, an idealised abstraction from data, these data being utterances (and inscriptions) in the language being described; but the idealisation is, standardly, a matter of methodological convenience, and properly recognized as such. It will be a matter for theoretical and metatheoretical debate what the exact relation is between data and system, or between the system, the linguistic community, and individual speakers.

The technical grammarian, in contrast, in assuming analogy to constitute the true or proper nature of (one part of) language, was thereby constructing, not a useful idealisation, but a genuine ideal, a model to be emulated. Any genuine departures from regularity were seen as forced concessions to a determinedly irregular and unideal world inhabited by imperfect and irrational creatures. That these may have been wise concessions, given the way of the world, is, to be sure, important for the history and social context of grammar, but it is theoretically irrelevant: the point is that such irregularities were perceived as aberrations from a perfect state of language existing independently of the people who use it (cf. von Fritz, 1949: 349). From the perspective of the technical grammarian, ordinary speakers, simply in virtue of being ordinary speakers, use a kind of language which may well be inherently irregular, and hence inferior compared with the ideal in which the grammarian is expert; and it is inconceivable that anyone could be an expert, a [GREEK TEXT NOT REPRODUCIBLE IN ASCII], in something which of its very nature is irreducible to rule.

Now it may be argued--was argued--that ordinary language only looks irregular: on closer inspection, apparent irregularities at any leve--lphonological, orthographic, morphological, syntactic--may turn out to be covered by some further rule, with all these rules falling under appropriate general principles. It seems (as already observed) that Apollonius Dyscolus attempted some such scheme in his Syntax, at least for received literary usage ([GREEK TEXT NOT REPRODUCIBLE IN ASCII]), and the method of `pathology' came to be widely used and respected, for morphology if not for syntax (see Blank, 1982: 41ff.).

Admittedly, again, certain expertises were held to exist in which successful performance is not guaranteed, not through fault of the expert, but because the matter on which the expert(ise) acts is inherently variable. The doctor, for example, may follow all the correct diagnostic and therapeutic procedures, arid still lose his patient, because individuals differ in their constitution, and, consequently, in their response to both illness and medical treatment. No-one challenges the doctor's claim to a certain sort of specialized knowledge just because not all his patients get better. Could there be some comparable escape clause for ordinary speakers? Perhaps their knowledge of their language is `stochastic' too: perhaps not all their linguistic activity need be correct for a claim to technical expertise to hold good. The material to which speakers apply their knowledge may be, like the doctor's, inherently variable, and so ultimately outside their, or anyone's, control.

I am not sure that we can push this analogy much further. For one thing, it is not clear, at least to me, what would count as a successful technical performance in a speaker, or what the material is on which he exercises his expertise. It is surely significant that we seem to have no ancient sources to guide us in these reflections. But in any case such manoeuvres will profit the ordinary speaker nothing: if anything, they make his claim to knowledge, at least knowledge of a sort to interest the grammarian, still more hollow.


The failure to recognize speakers' mastery of the language system, even if mastery in an incomplete or imperfect form, is surely to be accounted for by the grammarians' own epistemological assumptions. Here we must make use of and interpret the same concepts and models of knowledge to which the grammarians had access, and which were taken over, at least in part, from the epistemological professionals, the philosophers. In the course of examining, assessing, and applying the lessons of ordinary usage, grammarians focussed exclusively on what was said by users, without, it seems, being impressed by, or interested in analysing or exploiting, the cognitive resources which made that usage possible.

Reference has already been made to the grammarians' anxiety about the epistemological status of their discipline. It has been claimed that the (putative?) quarrel between analogist and anomalist grammar had its origins in disputes between rationalist and empirical doctors, and is to be understood as one special form of a general epistemological debate between champions of [GREEK TEXT NOT REPRODUCIBLE IN ASCII] and those of [GREEK TEXT NOT REPRODUCIBLE IN ASCII] (cf. Siebenborn, 1976: 8f., 1 1 6ff.). This is not the place to enter into the complex debate over grammar's self-image. The point to note is that--even if it is unclear how a speaker's cognitive state would have been characterized--ordinary users would certainly not have qualified as masters of any [GREEK TEXT NOT REPRODUCIBLE IN ASCII], even of their own language, let alone that of grammar. In particular, not only have speakers undergone nothing in the way of formal and explicit training, but also they cannot produce on demand statements, couched in appropriate jargon, of the general rules governing the properties of their usage. They will lack too even the most basic technical metalinguistic vocabulary, except in so far as a grammatical education has given them access to one even though they are perfectly well able to talk about words and their complexes, as speakers commonly do, in order to request or pass on information about pronunciation, meaning, or idiomatic usage. (It is interesting to observe that the explanation for Chrysippus' famous dictum, `every word is ambiguous' (Aulus Gellius, noct. att. XI xii 1), may be a claim that all words signify themselves as well as something else (see Atherton, 1993: 298ff.): one wonders whether he had been struck by this aspect of ordinary users' talk about language). A fortiori, speakers are also unable to judge whether or not this or that rule of language is a `good' rule, one which follows some principle commendable on independent grounds--say, its contribution to semantic or pragmatic efficiency. This principle is of great importance in Apollonius Dyscolus' theory of language: a frequent argument with him is that this or that grammatical feature has its roots in avoidance of ambiguity (e.g. synt. I 80, p. 68.2ff. Uhlig; and see Atherton [1995]).

It is far from obvious even that speakers have `experiential knowledge', [GREEK TEXT NOT REPRODUCIBLE IN ASCII], as, say, an empirical doctor has--his own carefully collected and organized experiences of particular cases. Certainly, users' knowledge has not been supplemented by wide reading of the relevant literature, and they have no access to a consciously articulated and consciously applied principle of `transition to similar cases' when attempting to produce or understand a new utterance. (On medical [GREEK TEXT NOT REPRODUCIBLE IN ASCII]see Frede, 1987c, esp. 245-7; on medical [GREEK TEXT NOT REPRODUCIBLE IN ASCII] and grammar, Siebenborn, 1976: 123-39.) Speakers cannot produce anything like the requisite generalizations about the language they speak, whether they refer to hidden entities and hidden causes, or prefer to stick to what can be observed. It is equally irrelevant, from this point of view, whether reason is given a role in scientific inquiry, whether what speakers may know (and grammarians prescribe) are natural or artificially-created arid sanctioned rules, and, finally, whether these rules are exceptionless or hold only for the mostpart', [GREEK TEXT NOT REPRODUCIBLE IN ASCII]. Probably the best the ordinary user can lay claim to is mere `experience', [GREEK TEXT NOT REPRODUCIBLE IN ASCII], of which Apollonius Dyscolus, for example, speaks with disdain, contrasting it with the understanding to be had of the reason, logos, behind this or that grammatical phenomenon (synt. I 60, p. 51,11. 1-7 Uhlig), but which Sextus commends as preferable to allowing oneself to be ruled by analogy (M. I 190, `above all the man who is practiced and experienced [GREEK TEXT NOT REPRODUCIBLE IN ASCII] in ordinary usage will speak good Greek, and not the one who understands analogy'). In any case, grammarians clearly did not think it worth investigating what sort of cognition speakers have. If we want illumination on this problem, we might try the philosophers, who of course have a long tradition too of worrying about types and degrees of knowledge: but that is another story.


Our whole discussion hitherto of the criteria of linguistic correctness has begged one crucial question: we have assumed that there is such a thing at all as correctness in language. It is because speakers do not, or so it is claimed, speak a correct language, and cannot give a technical account of the rules of the language they do or ought to speak, that they do not merit the technical grammarian's attention, and need instruction from him. But it may be questioned whether the grammarian is right to take for granted even so much as the fact that some usages are correct and some are not.

It is no answer to this question to point to the radical and wide-ranging changes which Greek and, later, Latin underwent as a result of political and social upheavals. For in principle technical grammarians might have regarded the state of everyday language at any one time and in any one community as the sole object of their inquiries, or they might have taken into account variability in time, space, and social stratum, while classifying all variations from a given usage as just that--variations. Social change and social pressure--notably, pressure from above to conform to `better linguistic habits could well explain why these rather than those usages came to be regarded as inferior or superior: but the question is rather why this or that sample of discourse can be taken as correct, or incorrect, rather than as merely different from some preferred model. Why should correctness or purity, rather than, say, success by some reference to a given purpose or function of language, ever come to our attention?

It might be suggested that we are dealing simply with an historical accident an unplanned, unforeseen effect of military conquest and political expansion, which meant, more and more as time went by, that Greek and Latin had to be accessible to speakers of other languages. After all, one principal offence against purity was called, precisely, `barbarism', the sort of error foreigners might make, under the influence of the sounds, formal patterns, idioms, and so on, of their own language (e.g. ad Her. IV 17; Cicero, de or. III 48; Quintilian I v 8 distinguishes this as just one type of barbarism more broadly conceived). Is this, then, the source of the concept of linguistic correctness: the actual or feared intrusion of elements of one language into another? Such cross-linguistic corruption was not confined to Greek: at least, Herodotus IV 117 refers to people `solecising in Scythian', [GREEK TEXT NOT REPRODUCIBLE IN ASCII]. This second major species of incorrectness, `solecism', itself supposedly derives its name from a colony whose isolated inhabitants had grown apart linguistically from their mother-city: but difference, again, is not incorrectness--unless one unthinkingly yields to parochial prejudice.

Was the recognition of correctness by grammarians perhaps the result of another accident: the phenomenal growth of the education industry in the Hellenistic period? Literary classics unintelligible even to native speakers had to be taught using `correct' texts and according to linguistic rules inapplicable to contemporary language. (Grammarians were quick to impress on their pupils that what is a fine poetic stroke in Virgil is mere impertinence in a schoolboy: cf. Bonner, 1977: 200; Kaster, 1988: 195.) But, again, the fact that the rules of this or that literary dialect or idiolect do not fit another version of Greek or Latin does not make the latter `wrong' in some way. An editor of a text might regard certain readings as inconsistent with the rest of his text, and so as less acceptable, without supposing that it would be incorrect to adopt them. The same can be said of synchronic differences between dialects. As for orthography, variations in spelling might be regarded as no more than that: surely the proper view here is that one spelling can be more easily intelligible, and so more useful, than another, and that a certain amount of consistency is also a useful thing (cf. Sextus M. I 171f).

Clearly these are not the views of ancient language theorists. It cannot be denied that difference just was typically seen by the grammarians as degeneration, across time, space, and social class, and the existence of correctness and incorrectness in language was never questioned. We ourselves stand at the end of millennia of debate about linguistic correctness, and a century or two of descriptive linguistics: yet the beliefs that some usages are inherently `good', others `bad', and that the language is constantly changing for the worse, are as deeply rooted now in the popular imagination as they were in the minds of ancient grammarians. Today's contributors to what Milroy and Milroy call `the complaint tradition' may not even think it necessary to adduce reasons for their criticisms (cf. Milroy and Milroy, 1991: 38). They assume, that is, not only that usages can be correct or incorrect, but that some are obviously one or the other. Ancient grammarians, it must be conceded, did not fall prey to this sin of omission. But they lacked one immensely useful distinction adopted almost universally by linguists today: that between language system (a concept discussed earlier) and use of that system. That this distinction has not found its way into the public domain today means that it is common for complaints about incorrectness or grammaticality, and complaints about ineffective, inappropriate, or unclear language, to become fused (cf. Milroy and Milroy, 1991: 37-40).

It would be unsafe to think of that now familiar distinction as shared also by ancient discussions of language, which some modern scholars have claimed was the case (e.g. Baratin and Desbordes, 1987: 53). But we have also had occasion to notice a very widely used ancient distinction which is comparable in some ways to the modern dichotomy: that between the linguistic excellence of purity or correctness, on the one hand, and, on the other, the remaining excellences of language, most commonly clarity, appropriateness, and elegance. The parallel is not exact; for example, factors which would be today excluded as irrelevant to the system, such as the origin of words, or literary authority, were standardly invoked to prove correctness. Likewise, even if formal analogy is accorded special status, as being internal to a language, semantic and pragmatic features could still play a role in explaining departures from regularity (e.g. Varro, L.L. IX 56).

It is at this point that Sextus Empiricus' onslaught on technical grammar proves its value, not only as a guide to the weaknesses inherent in the analogical programme, but also in reminding us of the real origin of the notion of correctness. In his critique, of the notion of hellenismos in particular, Sextus urges the merits of `usage' over all other criteria of linguistic purity or correctness (M. I 176ff.). The crucial feature of Sextus' strategy is already to be found in Varro's summary of arguments in favour of anomaly as a principle of grammar (L.L. VIII): the appeal to utility. Correctness has been found to be useful (note the typical Skeptic practice of falling into line with proven practical utility) only insofar as it permits a speaker, first, to avoid the social embarrassment of committing barbarisms and solecisms, and, second, `to present his thoughts about things both clearly and precisely' ([GREEK TEXT NOT REPRODUCIBLE IN ASCII] M. I 176). Such hellenismos, it is claimed, has nothing to do with `grammatical analogy', but rather `is derived, by reference to the usage of each Greek speaker, from parallel word-formation and from observation in the course of ordinary discourse' ([GREEK TEXT NOT REPRODUCIBLE IN ASCII][sections] 176), and because usage is `anomalous and inconstant', analogy, which arises from usage, has the same properties too, and has no fixed rules ([sections] 236). Its usefulness has already justified the existence of, and the need for, the expertise of [GREEK TEXT NOT REPRODUCIBLE IN ASCII], which provides instruction in basic `R and R': everyone, even Epicurus, who derided the traditional literary education (cf. [sections] 1), agrees on this point ([sections] 49). Indeed, the very arguments which might be levelled at the usefulness of this discipline, as they could not be remembered or passed down to later generations without it, are actually self-refuting ([sections] 53).

The unusual justification Sextus offers for seeking linguistic purity and elegance deals yet another blow to stylistic orthodoxy ([sections] 194). Sextus claims that `clarity and agreeable presentation of the things being indicated', [GREEK TEXT NOT REPRODUCIBLE IN ASCII], are the two principal items which justify concern with linguistic correctness, `for joined to these, externally and consequentially, is discourse using metaphor, emphasis, and the other tropes' ([GREEK TEXT NOT REPRODUCIBLE IN ASCII]). In other words, two standard excellences'. purity and elegance, are being subordinated to a third--which itself takes an unusual form. `agreeableness' being conjoined to the orthodox clarity'. Sextus has in view, I take it, a performance whose success depends not only on the communication of a message, but also on acceptance of it by one's audience. We may contrast here the Varronian anomalist, who allows refinement', elegantia, a role distinct from that of utility in the defence of irregularity in language (L.L. VIII 31f.).

Sextus' manoeuvre is accomplished tacitly, with no fanfare and no argument: but looking elsewhere we can easily see why he might have thought himself justified in demanding such a justification of Hellenism. For the assumptions underlying this portion of his critique of orthodox `good Greek' are ones shared by at least the bulk of ancient theorizing about language. According to the orthodoxy, a distinction can and mast be made between content and medium. Often, this content is itself glossed as `thought', the function of language then being the conveying of thought from one mind to another; and successful language is language which conveys a message, or does so best. This model of language is, in one form or another, so pervasive, so dominant, in ancient theorizing about language that it is often invisible, and almost always unquestioned. Sextus himself need not, of course, be endorsing this (or any) model of language; instead he exploits it for his own polemical purposes, in this case to use the orthodoxy about linguistic success to overturn the orthodoxy about linguistic purity.

For if `good' language is language which gets one's message across, then all other qualities must be subordinated to that aim. To achieve that end one's audience must be taken into account, in part--the part Sextus concentrates on--by using words they themselves use and understand: after all, they alone can be the judges of communicative success. As a result, correctness will be justified only insofar as it contributes to, is partly the means of, conveying something to others. It has no independent value, and, if it poses a threat to communication, it must be abandoned ([sub sections] 195f.). In effect, all linguistic activity must be judged by its communicational effectiveness, and the criterion of all competence with regard to language will be, so to say, its success in the field. This is also the assumption, for example, at [sections] 318f., where Sextus is arguing that the art of grammar is useless for exegesis of poetry `since according to them [sc., the grammarians], the best poem is the clear one; for clarity is poetic excellence, and inclarity, according to grammar, a defect' ([GREEK TEXT NOT REPRODUCIBLE IN ASCII]). In consequence, exegesis of the good (= clear) and of the bad (= unclear) poem alike is uncalled for. Sextus' complaint is unconvincing as it stands--he implicitly grants that clarity and its opposite are not, in grammatical orthodoxy, the sole poetical excellence and defect respectively--but he is, I take it, challenging his opponents to defend their aesthetics while maintaining their commitment to a model of language as a medium for a single, separable message.

It hardly need be said that this functional model of language is itself wide open to criticism. But positive theorizing is not Sextus' goal. His aim is discomfiting the opposition, and attacking implicit assumptions is as effective a method as any. No orthodox ancient writer on stylistics would have disputed the fact that clarity, among the other principal excellences of discourse, was a relative matter: intelligibility will be a function, not only of choice of vocabulary and sentence-structure, but also of audience, of context, of situation of utterance, and so on. What would be intolerably elliptical and vague in the court-room or lecture-theatre might be lucid and effective in the home. Appropriateness. too, of its very nature, is relative--to topic, speaker, audience. and situation or context. Type and degree of ornament will vary too, according to the perceived needs of the topic, the situation. and so on. Precision in one context would be pedantry in another: a murder charge is not to be argued as a case about rights of way might be; and so on. Even the Stoics, least orthodox of rhetoricians and orators, seem to have accepted that stylistic ornament, like clarity, is relative to subject-matter, although concessions were not made, in contrast, to audience or speaker or context: see D.L. VII 59 (with Atherton, 1988; 1993: 89f.), and. for a more conventional view, Cicero, de or. III 211f .

Given that several criteria of linguistic purity were accepted and applied, it would seem proper for grammarians to have accepted that correctness is also, in a sense, relative, as we saw earlier. But grammarians would be far less concerned to cut their linguistic cloth to suit their audience than were rhetoricians, who had the practical goal of success in the public arena always before them. Allowances must be made for dialect, of course (e.g. Quintilian I v 29, and note his comment, at I v 8, on Catullus 97.6), and appeal can be made to the authority of the past (as at I vi 11) or of a respected author. From the perspective of the theorist of purity, however, a speaker's (or writer's) audience is of comparatively little weight; and it is of no weight at all in deciding whether a given form is analogically or etymologically correct. Sextus, in contrast, is fully alive to the demands of context and situation. He is clear that we should take into account what our audience expects, and this will differ from audience to audience--social and educational levels in particular will be of crucial importance. In such cases, we should hear and obey: we should patiently submit to correction from the people whose language we are trying to use correctly ([sub sections] 191f):


For in everyday conversations the common run of people will either oppose us over certain expressions or they will not oppose us. If they do oppose us, they will at once correct us too: so that good Greek is to be had from those in ordinary life, and not from the grammanans. But if they do not disapprove of us, but understand the things we say, as being clear and correct, we too will abide by them.

How else, after all, can we be `clear and acceptable' to people, but by using their own language?

This, finally, is the nub of the matter. If we ask: why must we associate correctness with language, it is surely only because correctness emerges directly and inevitably from ordinary, formally untutored, linguistic behaviour because one of the activities in which speakers always learn to participate, and one by which they themselves learn to use a language and extend their grasp of it, is correcting themselves and each other. Language learners--not just `barbarians', but also the children within a linguistic community, acquiring their native tongue--will encounter any number of corrective utterances, unreflectingly given and often just as unreflectingly received and put into practice. Speakers ask and tell each other how to pronounce such-and-such a word, or when to use such-and-such a phrase, and whether it is all right for them to use such-and-such a sentence in such-and-such circumstances. Admittedly, such corrective activities are so pervasive today, at least in the West, because of the continuing influence of prescriptive grammar in education, and even in popular entertainment (we cannot play Scrabble or Call My Bluff, or do crosswords. without a dictionary, and a national dictation competition is held in France each year), and that influence must be explained by broad social and political tactors. But Sextus reminds us that speakers correct each other naturally', as it were, just as a function of being speakers, who come to be able to wield authority over learners in part precisely because they have learned both to ask questions and to adapt their practices to the answers given.

Of course, this does not mean that speakers must have a firm metalinguistic concept of correctness, as opposed to. say, one of acceptability--which is another point to which Sextus draws our attention. The grammarian or linguist will distinguish narrowly linguistic features (grammaticality. correctness) from broader features of language use (where preferences are based on clarity, appropriateness, elegance, and so forth). Aristotle seems to have had both a narrower notion of `pure Greek', an offence against which constitutes a solecism, and a broad notion of `Hellenism' or good Greek' (contrast s. el. 3 165b20f., 14 173b17ff. with rh. III v 1407a19ff.), but it took time for the two to be systematically distinguished (see Innes, 1985). As this is a technical distinction, made for theoretical or methodological or paedagogic purposes, it is, however, all too easy to forget purity's humble origins.

For if there is one great mystery surrounding language learning and language n knowledge, it is, as we noted much earlier, that the rules which grammarians and linguists compile are manifestly not ones immediately, intuitively accessible to the ordinary speaker. The rules which people intuitively appeal to for guiding and correcting themselves and others are, I think it can safely be said, never the rules d which grammar constructs, even if both sorts purport to govern the same sort of utterance. Users' rules are not abstract and general; users themselves lack all but the most basic, non-technical, metalinguistic vocabulary; they will tend to appeal to examples, whether explaining an idiom or showing how to use or not use a given word or phrase; and they will do all this often without so much as noticing that they are t talking, and in quite general terms, about' their language. They can teach someone an inflection or a construction, and they do not even need jargon to do so (`say taught not leached'). An ordinary English speaker will tell you readily enough that *`Man the came into room the' is wrong, as is *`Did you came yesterday?' end

*`These are the houses that we didn't know what were like inside', and, what's more, will give you the corresponding correct sentences; but will have nothing to say about the position of definite articles, or interrogative transformations, or topicalized sentences containing relative clauses with embedded clauses (I owe the last of these examples to Milroy and Milroy, 1991: 78).

What is more, speaker's rules can be regarded as `rules' in a quite different sense from the `rules' of linguistics. For the source of the concept of linguistic correctness--as opposed to that of regularity--must be users' own intuitive activity of correcting each other and themselves. If it were not for speakers teaching each other what can be said and what cannot, all the grammarians' propositions about the languages they describe would surely have no more normative force than have the meterologists' propositions about the `normal' rainfall for England in July. The crucial difference is between something happening `as a rule', and something happening because of a rule (although of course rules do not cause behaviour: see Baker and Hacker, 1984: 257-9, and indeed the whole of their discussion of rules (pp. 243ff.), to which I am much indebted).

Apparently, however, all this was not a puzzle for the ancient grammarian. Ordinary people so obviously lack technical knowledge of their language (which is, in part, articulate knowledge of appropriate technical concepts and jargon) that the question seems never to have arisen, how they manage so much of the time to produce what must be admitted to be correct utterances. It is today a commonplace that a language is a system of rules, even if there is no guarantee that it is either a genuine system (a set of rules which is comprehensive, coherent, and organized) or a stable one: it is also a common-place now (as it seems to have been in Varro's day, L.L. IX 17; and cf. Horace, ars poet. 60ff., Sextus, M. I 82) that ordinary language is always changing, so that descriptions purporting to represent static and unvarying systems must be recognised for the idealizations they are. It is often argued too that speakers must somehow `know' the rule system, and just as heatedly denied that this is possible, or even coherent. But we do not find in antiquity, at least in ancient grammar, either any form of that problem, or a distinction which has in modern times been brought to bear on it, that between doing something according to rules, and doing something which can be described by rules.

Sextus does talk about the `rules' ([GREEK TEXT NOT REPRODUCIBLE IN ASCII]) of grammar: but nowhere does he even suggest that what ordinary speakers know, as opposed to what they say, should serve as the criterion of what is or is not good Greek. Sextus' own philosophical position alone might seem sufficient reason for this silence. As a Skeptic, he would never venture so dogmatic a claim as that speakers, any more than grammarians, `know' anything at all--at least according to available doctrinaire conceptions of what constitutes `scientific knowledge', [GREEK TEXT NOT REPRODUCIBLE IN ASCII], or `expert knowledge', [GREEK TEXT NOT REPRODUCIBLE IN ASCII] so that all ascriptions by Sextus of something approaching `knowledge' must be treated with caution. I take it that he is adopting (appropriately enough) the ordinary, non-technical, non-doctrinaire, usage of such words as `know' and knowledge', without thereby committing himself to the belief that speakers do genuinely `know' anything, in any sense (cf. Frede, 1987c: 255; Barnes, 1983: 154f.). Thus at [sections] 44 grammar is described as `in the general sense, knowledge of letters of any sort whatsoever, [GREEK TEXT NOT REPRODUCIBLE IN ASCII] at [sections] 50,52 it is allowed that [GREEK TEXT NOT REPRODUCIBLE IN ASCII] is an expertise; Athenians have [GREEK TEXT NOT REPRODUCIBLE IN ASCII] of `the expressions customary in Attica', [sections] 87; even grammarians are said `to have learned', [GREEK TEXT NOT REPRODUCIBLE IN ASCII], the meanings of foreign and unusual words--not from any expertise, mind you, but from other people, the users of these words ([sections] 313); and at [sections] 61 Sextus explicitly appeals to ordinary usage of the word [GREEK TEXT NOT REPRODUCIBLE IN ASCII], experienced.

In fact, ordinary speakers, whose usage, [GREEK TEXT NOT REPRODUCIBLE IN ASCII], is not a technical matter ([sections] 219), seem to be conceded no more than [GREEK TEXT NOT REPRODUCIBLE IN ASCII], `experience': `Greeks and barbarians alike, plain men and cultured men alike', will hear the `vocal sound' ([GREEK TEXT NOT REPRODUCIBLE IN ASCII]) when it is utterea, `but discourse and its parts [GREEK TEXT NOT REPRODUCIBLE IN ASCII] are understood only by Greeks, that is, by those with experience of this <discourse>' ([GREEK TEXT NOT REPRODUCIBLE IN ASCII]); this last phrase could perhaps be understood as referring to `Greeks and people skilled in Greek', the implication being that a native speaker's knowledge of Greek is something different from knowledge of it as a second language ([sections] 155). (There is an art of conversation, [GREEK TEXT NOT REPRODUCIBLE IN ASCII], it seems, but this too has its basis in `a sort of general aptitude', [GREEK TEXT NOT REPRODUCIBLE IN ASCII] ([sections] 295)). To be preferred to any so-called grammatical expertise is the Skeptic's own non-dogmatic, non-theoretical, adherence to `observation', [GREEK TEXT NOT REPRODUCIBLE IN ASCII], of practices in the ordinary course of every day life, [GREEK TEXT NOT REPRODUCIBLE IN ASCII]: and the suggestion is that all the skeptic does is observe what common people do, and copy them. Thus he will abide by `a sort of everyday, plain usage of ordinary people' ([GREEK TEXT NOT REPRODUCIBLE IN ASCII]), which differs from city to city and people to people ([sub sections] 232), and stick to `observation [GREEK TEXT NOT REPRODUCIBLE IN ASCII] which is non-technical, plain, and everyday, and accords with the common usage of ordinary people' ([sections] 179), just as Homer followed the usage of his contemporaries ([sub sections] 207f.). It is important to Sextus' skeptical strategy that language prove a matter of convention (e.g. M. I. 143ff., with VIII 134, XI 241f., P.H. II 214, III 267), not least because correctness, like everything else about language, can then be presented as socially determined, as no more natural than are fashions of clothes or table-manners: correct usage is neither natural, nor technical, but a mere collocation of arbitrary customs (cf. M. I 187, 236).

One of Sextus' repeated criticisms of grammar itself is that it lays claim to knowledge (broadly understood) of an indefinite, [GREEK TEXT NOT REPRODUCIBLE IN ASCII], number of items, whether the language of the poets, of all literary writers, or of ordinary people (e.g. [sub sections] 65ff., 221ff., 259, 314). The obvious response (cf. Itkonen, 1991: 202) must be that the grammarian has grasped a definite set of rules which serve to organize and explain this apparently unwieldy and indefinitely large subject-matter. (This is most apparent in the case of technical grammar, but even the grammar which deals with correction and exegesis of texts can lay claim to principles and rules applicable to any appropriate subject matter.) There is a world of difference between an indefinitely large, and an inherently irregular, set of objects. But the closest Sextus comes to tackling this tricky point is a complaint that even the rules, [GREEK TEXT NOT REPRODUCIBLE IN ASCII], which grammarians construct do not apply to every word, [GREEK TEXT NOT REPRODUCIBLE IN ASCII], but only to, say, certain types of lexical prosody, and thus have a strictly limited field of application ([sections] 89). It is significant that he turns at once to examine the subdivisions of grammar ([sub sections] 91-6), and then to detailed criticism of the technical part ([sub sections] 97ff.)--precisely the part of grammar which produces such rules as these.

One of the most striking features of Sextus' attack on grammar is his concentration on analyses and explanations of the forms and properties of individual words or parts of speech. Syntax is remarkable by its absence most notably at [sub sections] 132-40, where Sextus sets out to show that the first line of the Iliad does not exist, at least as analysed by grammarians into the standard parts of speech. This famous logos (and, by implication, any word-complex) is presented as a mere `aggregate', [GREEK TEXT NOT REPRODUCIBLE IN ASCII] ([sub sections] 134, 135, 136), of its parts: indeed, Sextus' complaints seem so puzzling precisely because they ignore linguistic structure almost completely (much as do the puzzles about letters and syllables in Plato's Theaetetus, 204b-205b). Only at [sections] 137, which mentions `each [of the parts of speech] having its own proper place', [GREEK TEXT NOT REPRODUCIBLE IN ASCII] is there a hint that sentences might have internal structures, at least by way of linear ordering. Technical grammar itself is said to embrace just `the elements and the parts of speech, orthography and Hellenism, and connected matters' ([sections] 92), and, similarly, the detailed treatment of it ([sub sections] 97-247) hardly touches on syntax at all.

Now it has been claimed (e.g. for the Latin tradition, see Charpin, 1978; Collart, 1978a: Baratin, 1989: 7f.) that ancient grammarians were, as a rule, not interested in syntax, which was the Cinderella of technical grammar, the ugly sisters being prosody, orthography, the theories of the letter-sounds, syllables, and parts of speech, and so on; so perhaps Sextus' neglect of syntax is no more than a reflection of his opponents' interests. But, first, there is no reason to expect constant even-handedness from Sextus. He is no doxographer; his role is primarily that of critic, not reporter; and his criticisms might be just as effective if veiled, indirect, even deliberately obtuse. We may compare his criticisms elsewhere of the dialecticians' treatment of ambiguity, which might also be expected to touch on syntactic ambiguities, and yet do no such thing, although we know from other sources that syntactic ambiguity was of as much interest to dialecticians as the lexical variety (P.H. II 256-9, with Atherton, 1993: 86f., 430). Second, it may be rash to infer from the relative scarcity of treatises dealing explicitly with syntax, such as those of Apollonius Dyscolus and Priscian, that syntax was ignored. Rather, it might have been handled in a different way, under the rubric of the individual parts of speech (Barwick, 1957: 25f. argued that this was the Stoic approach); and in fact Apollonius' Syntax, the single most sophisticated and original work on the subject to survive, approaches it in much this fashion (see, e.g. synt. I 36, pp. 33.9-34.2 Uhlig). So Sextus' silence on syntax might turn out to be, rather, a reflection of his opponents' choice of method.

What is more, the almost casual presentation of the paradoxes about logoi could have been intended to provoke suitably splenetic responses from grammarians, especially if--as would almost certainly have been the case they contradicted each other. That is the scenario envisaged by Sextus, for instance, in the case of the dialecticians' handling of ambiguities (P.H. II 259). Note, too, that for Sextus their treatment of ambiguities is as unnecessary as it is groundless: ordinary people, even slaves (the same slaves as master inflections, according to Varro), can cope with ambiguities perfectly well in the everyday course of affairs. while, in areas of special expertise, such as medicine, we turn to recognised experts for guidance, not to dialecticians.

Yet if syntax was. typically, dealt with neither as a separate subject, but as one of the properties of each of the parts of speech, nor systematically, but rather as and when needed--say, when one had to deal with solecisms or other problems (e.g. Apollonius synt. III 6, pp. 271.5-272.3 Uhlig; and cf. Blank, 1982: 6ff.)--then the place of syntactic rules in linguistic knowledge might tend to be suppressed. It has today become something of a commonplace that language users can say and understand things they have never said or heard before: whatever it is they know, it governs or embraces. not a finite set of utterances, but an indefinitely large number of them; and their creativity is often taken to lie in their mastery of a creative (generative) system. A modern linguist might want to claim that ordinary, untutored users have mastered a set of rules, rather than that, less controversially, languages must be describable as governed by finite sets of rules capable of producing an indefinitely large number of sentences. Long ago von Humboldt described language as `making infinite use of finite resources'. Varro arrived at a similar conclusion even earlier--but, as we saw, he made very little of it, and other grammarians seem not even to have got so far. And if one thinks of knowing syntax as like knowing how to pronounce a word, or what its cases are--just knowing one more thing about it that knowledge will look far less impressive and remarkable than it actually is.

Sextus, too, shows no sign of recognizing the marvel of ordinary language knowledge for what it is. Glidden (1983: 241ff.) has argued that Sextus nowhere reveals his own beliefs about language and language learning, and even rejects elsewhere the account of how languages are learned--by appeal to commemorative signs--which he might appear to espouse here (the difficulty is that language cannot be reamed without learning what it is that words, like all signs, stand for [cf. M. I 37f., P.H. III 267f.]). (Desbordes, 1982: 53f. claims that M. I displays a non-Skeptic partipris against grammar, but Glidden's reading of the text does far more justice to Sextus' Pyrrhonism.) Sextus is (sometimes) careful to allow that different usages hold sway in different circumstances and contexts, and allows quite happily that the skeptic will follow whichever one is appropriate--whether that of the pot-boy, the doctor, or the philosopher ([sub sections] 232-5), of the Athenian, the Thracian, the Roman, or the Spartan ([sub sections] 218, 228; cf. 187, 213, and especially [sections] 89: an indefinite number of idiolects exists within each dialect). Even the analogist grammarians must admit that there are different usages having different regularities ([sub sections] 229-31). Elsewhere Sextus speaks more broadly as it suits his purpose of `the common usage of all' or `...of the Greeks' or `...of each of the Greeks' ([sub sections] 97, 176. 179, 188, 193).

What is deeply unconvincing is his suggestion that what the skeptic is doing, in falling in with ordinary usage, is merely doing what he is told to do on a particular occasion (M. I 191f.). For the skeptic will himself learn, or master, or have learned or mastered, the conventions particular applications of which he obeys; crucially, he will learn to apply them on unfamiliar occasions to produce or understand novel and unfamiliar utterances. `Following the usage of the many' ([sections] 193) is not a matter of making particular choices about the use of a given token in this or that particular circumstance. The skeptic will also learn how to use the word-type in other, similar circumstances in future, without further coaching. that is, the skeptic himself will surely acquire whatever linguistic resource it is that slaves and philosophers alike have recourse to, and give evidence of, when they speak without consulting anyone, and, equally, when they correct their own or another's usage.

So it is not enough to assert that the skeptic will master a finite list of (types of) words correlated with a finite list of (types of) circumstances, forming a matrix from which the right word can be mechanically selected and used as required. The skeptic must master something which allows him to understand and produce an indefinitely large number of utterances acceptable to speakers of the dialect or subdialect to the norms of which he is attempting to conform. If the skeptic does not do that, then despite his protestations he is ipso facto not doing what ordinary people do when they correct themselves and each other. What is more, merely by claiming to adhere to ordinary usage in a reflecting and rational fashion, Sextus is departing from the (ostensibly non-theoretical, unreflecting) position of the ordinary speakers he n pretends to emulate; and, although in describing usage as `non-technical' and `everyday', Sextus may evade the doctrinaire thinkers' demands for knowledge on their terms, he is, as always, choosing to adopt his opponents' cognitive schemes. Everything he says about language knowledge is to be construed in terms of the philosophical-epistemological debate between dogmatic and skeptic. If there is anything special about the cognitive basis of language use, about the process by which such cognition is acquired, or about its content and structure, it plays no part in the Pyrrhonist's wholly negative project. Sextus comes closest to forcing the grammarians' hand over ordinary language knowledge. But, in the end, he has no need to see them put their cards on the table, and he claims to have none of his own.


What, then, does every grammarian know--every technical grammarian, at least? He seems to know a great deal: about letters and their sounds and spelling, about syllables and their combinations, about parts of speech and their properties and relations. But one thing he has missed: the ultimate source of the very object to which he devoted so much time, energy, and ingenuity: correctness itself. By ignoring ordinary language use and activity, and whatever cognitive resources make that activity possible, the technical grammarian overlooked the fact, not that analogy has s its roots in usage, as Varro and Quintilian remind us, but that correctness does--specifically, in ordinary speakers' judgements about their own and others' utterances. The technical grammarian was wrong to think that the ordinary speaker knows nothing of any significance. For there is one thing which only the ordinary user, or the ordinary user in the grammarian, can show the grammarian: not, necessarily, that this or that bit of language is correct or incorrect, but that it is possible for this or that bit of language to be correct or incorrect. Of course, ordinary speakers cannot tell the grammarian why this or that is correct or incorrect (and may well disagree over the what and the when as well), and this ignorance, or inarticulateness, damned them in the grammarian's eyes. Nor can they tell him, by way of a claim true of language in general, why there is such a thing as correctness. But that was one thing the technical grammarians did not know either.(1)

(1) Stich, 1971 both inspired the title of this paper, and first suggested to me the reason for ancient grammarians' (and philosophers') neglect of language knowledge. My thanks go to Robert Wardy and to the anonymous reader for Classical Quarterly for their comments on this paper.



Atherton. C. 1988 `Hand over fist: the failure of Stoic rhetoric', CQ 38, 392-427. 1993 The Stoicks on Ambiguity (Cambridge). 1995 `Apollonius Dyscolus and the ambiguity of ambiguity', CQ 45, 441-73.

Baker. G. P., Hacker, P. M. S. 1984 Language. Sense and Nonsense. A critical investigation into modern theories of language (Oxford).

Baratin, M. 1989 La Naissance de la Syntaxe a Rome (Paris).

Baratin, M. Desbordes, F. 1987 `"La troisieme partie" de l'ars grammatica', in Taylor 1987, at 41-66.

Barnes, J. 1983 `Ancient Skepticism and causation', in M. Burnyeat (ed.), The Skeptical Tradition (Berkeley, Los Angeles, London), at 149-203.

Barwick, K. 1922 Remmius Palaemon und die romische ars grammatica (= Philologus Suppl. 15:2) (Leipzig; repr. Hildesheim, 1967). 1957 `Probleme der stoischen Sprachlehre und Rhetorik', Abhandlungen der sachsischen Akademie der Wissenschaft zu Leipzig, phil.-hist. Kl. 49, Heft 3.

Blank D. 1982 Ancient Philosophy and Grammar. The Syntax of Apollonius Dyscolus (Chico, CA: American Philological Association: American Classical Studies, 10.).

Bonner, S. 1977 Education in Ancient Rome. From the Elder Cato to the Younger Pliny (London).

Charpin, F. 1978 `La notion de solecisme chez les grammairiens latins', in Collart 1978a, at 211-16. Collart, J. (ed.) 1978a Varron, Grammaire Antique, et Stylistique Latine (Paris).

Collart, J. 1978b `A propos des etudes syntaxiques chez les grammairiens Latins', in Collart 1978a, at 196-204.

Desbordes, F. 1982 `Le langage skeptique', Langages 65, 47-74.

Ebbesen, S. 1981 Commentators and Commentaries on Aristotle's `Sophistici Elenchi' (Leiden, 3 vols.).

Frede M. 1987 Essays in Ancient Philosophy (Oxford). 1987a `The origins of traditional grammar', in Frede 1987, at 338-59. 1987b `Principles of Stoic grammar', in Frede 1987, at 301-37. 1987c `The ancient Empiricists', in Frede 1987, at 243-60. von Fritz, K. 1949 `Ancient instruction in "grammar" according to Quintilian', American Journal of Philology 70, 337-66.

Glidden, D. 1983 `Skeptic semiotics', Phronesis 28, 213-55.

Innes, D. 1985 `Theophrastus and the theory of style', in Fortenbaugh, W. W. (ed.), (with P. M. Huby, A. A. Long), Theophrastus of Eresus. On his life and work (Rutgers University Studies in Classical Humanities, 11, New Brunswick, Oxford), at 251-68.

Itkonen, I. 1991 Universal History of Linguistics. India, China, Arabia, Europe (Amsterdam: Philadelphia, Amsterdam Studies in the Theory and History of Linguistic Science, Series III: Studies in the History of the Language Sciences, 66).

Kaimio, J. 1979 The Romans and the Greek Language (Commentationes Humanarum Litterarum 74, Helsinki).

Kaster, R. A. 1988 Guardians of Language. the Grammarian and Society in Late Antiquity (Berkeley, Los Angeles).

Lallot, J. 1989 La grammaire de Denys de Thrace (Paris). Milroy, J. and Milroy, L. 1991 Authority in Language. Investigating language prescription and standardisation (London/New York, first publ. 1985).

Robins, R. H. 1957 `Dionysius Thrax and the Western grammatical tradition', Trans. Phil. Soc. 67- 106. 1976 `Varro and the tactics of analogist grammarians', in A. Morpurgo Davies, W. Meid (eds.), Studies in Greek, Italic and Indo-European Linguistics, offered to L. R. Palmer on the occasion of his 70th birthday (Innsbruck), at 333-6.

Siebenborn, E. 1976 Die Lehre von der Sprachrichtigkeit und ihren Kriterien. Studien zur antiken normativen Grammatik (Amsterdam).

Stich, S. P. 1971 `What every speaker knows', The Philosophical Review 80, 476-96.

Swiggers, P., Wouters, A. 1994 `The philosophical and ideological underpinning of grammatical and linguistic theories', Katholieke Universiteit Leuven, preprint no. 149.

Taylor, D. J. 1974 Declinatio. A study of the linguistic theory of Marcus Terentius Varro (Amsterdam: Amsterdam Studies in the Theory and History of Linguistic Science, Series III: Studies in the History of the Language Sciences, 2). 1987 ed., The History of Linguistics in the Classical Period (Amsterdam, Philadelphia: Amsterdam Studies in the Theory and History of Linguistic Science, Series III: Studies in the History of the Language Sciences, 46).

Versteegh, K. 1987 `Latinitas, Hellenismos, `Arabiyya', in Taylor, 1987, at pp. 251-74.
COPYRIGHT 1996 Cambridge University Press
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1996 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:the ancients
Author:Atherton, Catherine
Publication:The Classical Quarterly
Date:Jan 1, 1996
Previous Article:Finances, figures and fiction.
Next Article:A new fragment of the historian Theseus.

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters