You are on page 1of 18

The Development of the Word Class System of the European Grammatical Tradition Author(s): R. H.

Robins Source: Foundations of Language, Vol. 2, No. 1 (Feb., 1966), pp. 3-19 Published by: Springer Stable URL: http://www.jstor.org/stable/25000198 . Accessed: 20/09/2013 08:31
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Springer is collaborating with JSTOR to digitize, preserve and extend access to Foundations of Language.

http://www.jstor.org

This content downloaded from 198.168.27.222 on Fri, 20 Sep 2013 08:31:26 AM All use subject to JSTOR Terms and Conditions

R. H. ROBINS

THE DEVELOPMENT THE EUROPEAN

OF THE WORD GRAMMATICAL

CLASS

SYSTEM OF

TRADITION*

The twomost significantpredecessorsof modern descriptive linguisticshave been the work of ancient Indian scholarship in the fields of phonetics, grammar,and general linguistic theory, and the continuouswestern tradition of grammar established in all its essentials in ancient Greece. The formerbecame known to European scholarshiponly in the nineteenth century, though its influence thereafterwas profound.1 The latterhas been constantly a part of the cultural traditionof thewestern world. Greek theory was taken over by the Romans, handed on with but slightmodifications by the lateLatin grammarians (notablyDonatus and Priscian) to themediae val world, and, as a joint bequest from themiddle ages and the 'rediscovered' treasuresof classical antiquity, became the basis of language teaching and of much linguistic theorizing in European education and scholarship.
It has been decessors customary to stress the excellence of Indian work to the on them

disparagement of the Greco-Roman achievement. Panini and his pre


and successors in India well deserve the praise bestowed

by modern linguists2, and certainly contemporary phonetics and much of contemporary linguistic descriptive and analytical procedure, in particular
the theory of morphemic analysis, owes more to India than to the west.

It is, perhaps, a pity that the long traditionof western linguistic scholarship is judged by its relative failure in phonetics and its lack of any proper theoreticalbasis for etymology, rather thanby itsvery real success in devising a system for the grammatical description of Greek and Latin. The attempted imposition of this grammatical system on numbers of other languages un
related in structure to Greek and Latin is a charge against some rather

unimaginative moderns, not one against the Greek and Latin scholars whose interests and circumstances never led them to venture outside the twowestern classical languages, anymore than the ancient Indians analysed languagesother than Sanskrit. It is, incidentally, noticeable that among some transformationalists the
* I am indebted to Professor J. Lyons for kindly reading a draft of this paper and making a number of very pertinent comments. 1 See, especially on the Indian contribution to phonetics and phonology inEurope, W. S. Allen, Phonetics inAncient India, London, 1953. 2 E.g. L. Bloomfield, Language, London, 1935, 11: 'One of the greatest monuments of human intelligence'; cp. id., Language 5 (1929) 267-76. 3

Foundations of Language 2 (1966) 3-19. All rights reserved.

This content downloaded from 198.168.27.222 on Fri, 20 Sep 2013 08:31:26 AM All use subject to JSTOR Terms and Conditions

R. H. ROBINS

work of earlierEuropean writers within theGreco-Roman tradition is now credited with insights and understanding from which the descriptivists of the 1930s and 1940swere by their own approach to the subject deliberately excluded.3 The linguistic achievement of western antiquity was in grammar, in the narrower sense of that term, excluding phonetics and phonology; in these latter branches the relative poverty of phonetic observation and failure adequately tomaintain the distinction between lettersand sounds prevented really significantwork. But the descriptive grammar of Greek and Latin worked out in antiquity has survivedwith comparatively few alterations
for nearly 2000 years, and such alterations as have been made and accepted

can be accommodated in the systemwithout fundamental changes. of the classical tradition of grammar Perhaps the principal infrastructure the of of been has system parts speech (mere l6gou,partes orationis) or word classes and the grammatical categories (case, tense, number, gender, etc.) associatedwith them and serving to distinguish and define those classes whose members admit inflection.The Greek system of eight classes (noun, verb, participle, article, pronoun, preposition, adverb, conjunction) was established by Aristarchus (second century B.C.), and set out in the extant
grammar of Greek by Dionysius Thrax, his pupil.4 It was taken over un

changed as the basis of Apollonius Dyscolus's works on Greek syntax, and Priscian, expressly modelling his description of Latin on Apollonius's
account of Greek5, preserved the number of classes at eight, compensating 3 E.g. N. Chomsky, Current Issues inLinguistic Theory, The Hague, 1964, 15-27. 4 Text in I. Bekker, Anecdota Graeca 2, Berlin, 1816, 627-43, and G. Uhlig, Dionysii Thracis ars grammatica, Leipsig, 1883. The statements made in this paper assume the genuineness of the bulk of the text of the Techne grammatike as we have it.Doubts on its rightful ascription to Dionysius Thrax were raised first by the scholiasts (Bekker, op. cit. 672); arguments in favour of its being the genuine work of Dionysius Thrax were set out in L. Lersch, Die Sprachphilosophie der Alten 2, Bonn, 1840, 64-76, and more fully by M. Schmidt, 'Dionys der Thraker', Philologus 7 (1852) 360-82; 8 (1853) 231-53, 510-20 and these were accepted by H. Steinthal, Geschichte der Sprachwissenschaft bei den Griechen und Romern 2, Berlin, 1891, and Paully-Wissowa, Realencyklopadie 5.1, Stuttgart, 1903, s.v. 'Dionysius 134'. Recently the argument has been reopened by V. di Benedetto, 'Dionisio Trace e la techne a lui attribuita', Annali della scuola normale superiore di Pisa, serie 2, 27 (1958) 169-210; 28 (1959) 87-118; di Benedetto reexamines the earlier evidence together with recent discoveries of grammatical writings in papyri. In brief, his conclusions are that the text thatwe have from Section 6 on is a third or fourth century A.D. compilation tacked on to the introduction of a now lostwork by Dionysius. Whether or not this conclusion is justified, it remains the case that the system set out in the Techne is the one assumed by Apollonius Dyscolus (except in some details) and largely reproduced by Priscian, modelling himself on Apollonius (cp. the lists and definitions of word classes in the Techne, Section 13 and in Priscian, Institutiones grammaticae 2.4.15-21). The course of development and the relative chronology of the system examined in this paper are not affected by the question of textual genuineness. 5 E.g. Priscian, Inst. gram. 12.3.13, 14.1.1, 17.1.1. 4

This content downloaded from 198.168.27.222 on Fri, 20 Sep 2013 08:31:26 AM All use subject to JSTOR Terms and Conditions

THE

DEVELOPMENT

OF THE WORD

CLASS

SYSTEM

for the absence of a definite article in Latin corresponding to ho, he, to in Greek by the recognition of Latin interjections as a separate class of words, a step taken at least as early as the first centuryA.D. by Remmius
Palaemon.6

The importanceof theword class system in grammatical descriptionwas recognized in antiquity. A scholiast makes it Thrax's most important contribution7; and inmodern times one of the first systematichistorians of linguistics, Laurenz Lersch, takes the evolution of the classical eight class system as themain theme in his exposition of the development of Greek and Latin grammatical theory and statement.8 Panini's work on the grammar of Sanskrit, from its compactness and systematic economy (whence springs much of its difficulty) manifestly comes at the end of a long line of predecessors; but of these little is known. It is, however, possible in the Greek tradition to follow the growth of the Aristarchan word class system and observe the methods by which it was successively developed. It is clear from the form in which ancient writers put their statements that they saw the history of grammar as involving a word class system that was progressively expanded through the creation of new classes from the subdivision of classes previously recognized in earlier systems; and thismay be fairly represented in the diagram at the end of this
paper.

Some details remain in doubt or are the subject of controversy, but the resolution of these problems is amatter for classical scholarship rather than for general linguistics.The earliest attempts at a statementof Greek grammar are centred on the definition of sentence components, and the changes in their number and in themeans by which theywere established provide an interestingexample of progressive linguistic research, researchcarriedout by speakers of Greek on Greek9, and spread over several generations, but in some respectsnot unlike the course of development throughwhich progresses the analytic framework set up by a modern linguistworking in the field on a hitherto unanalyzed language.
6 Charisius, Ars grammatica 2.212. 7 Bekker, An. Graec. 2, 676: 6 Op&a AtovUatog, 6 iepi T-ov 6KTcogepcov Tob k6you &Sta;aq lgatq (Dionysius Thrax, who taught us about the eight parts of speech): cp. ibid. 724. 8 Lersch, Sprachphilosophie, part 2. 9 But it has been pointed out that significant advances and refinements in grammatical observation and analysis were achieved by the Stoics, whose founder Zeno of Citium, like his successor Chrysippus, was said to have learnedGreek as a second language (M. Pohlenz, 'Die Begriindung der abendlandischen Sprachlehre durch die Stoa', Nachrichten von der Gesellschaft der Wissenschaften zu Gottingen, phil.-hist. Klasse Fachgruppe 1 Altertums wissenschaft, N. F. Bd. 3 (1939) Nr. 6. On Stoic linguistics see further K. Barwick, 'Probleme der stoischen Sprachlehre und Rhetorik', Abhandlungen der sichsischen Aka demie der Wissenschaften zu Leipzig, phil.-hist. Kl. 49 (1957) Heft 3. 5

This content downloaded from 198.168.27.222 on Fri, 20 Sep 2013 08:31:26 AM All use subject to JSTOR Terms and Conditions

R. H. ROBINS

As the Latin grammar of Priscian was so largely modelled on theGreek from Thrax taken over and its theoretical basis was system by Apollonius the same, this studymay be confined in the main to the work of Greek scholars. Itmust be stressed that for the pioneers of Greek grammar,Greek was an unanalyzed language. It is hard to-day not to envisage Greek set out in the (largelyAristarchan) system of traditional grammar, and one is easily tempted to look on thework of Aristotle and the Stoics, the twomain exponents of grammatical systems before Aristarchus, as the progressive 'discovery'or 'revelation'of an already existent set of classes and grammati cal categories. There is no need to revive the controversy between what have been nicknamed 'hocus-pocus'and 'God's truth' attitudes to linguistic analysis, nor to ask the presumably unanswerable question whether gram matical structures and systems exist independentlyof a grammatical state ment. Certainly Greek and Latin were predisposed towards the grammar imposed on them in a way that some other languages to which the system was unimaginatively applied were not. In this sense Thrax succeeded
because Greek was 'like that', but we can hardly say that his was the only

model of statement or that his system of eight word classes was the only satisfactory one. It may have been the best, and certainly the subsequent modifications, involving the separate recognitionof the adjective fromwithin the noun class and the inclusion of his separate class of participleswithin within an existing the forms of the verb, are in thenature of a rearrangement framework and rely on the same defining categories; but prior to the estab lishment of the system therewill have been other ways inwhich a gram matical description of Greek could have been organized. The crucial de cisions that led to the adoption of the system set out by Thrax are seen to
have been taken as one follows One must Greek bear in mind to be discovered of his predecessors. was no preexisting grammar of as there that, just in the form in which we now have it, so equally the work

the purposes behind grammatical analysis changed along with changes in the conditions that fostered it. Plato was concerned with the structure of sentences as the vehicles of logical argument. Additionally, Aristotle set himself the task of classifying and defining the basic termsof the descriptive sciences in general. The Stoicsmade the study of language, including gram mar, etymology, and rhetoric, a central part of their philosophical investi gations. The Alexandrian literary critics saw grammar primarily as part of the equipment required for the appreciation of literatureand the estab
lishment of correct texts of earlier writers. From all of these points of

approach grammar grew up to become a scholarly activity in its own right, almost as a by-product of other objectives.
Prior 6 to the Stoics one can hardly speak of grammar as a separately

This content downloaded from 198.168.27.222 on Fri, 20 Sep 2013 08:31:26 AM All use subject to JSTOR Terms and Conditions

THE

DEVELOPMENT

OF

THE WORD

CLASS

SYSTEM

recognized discipline in the west; but questions of grammar are discussed by both Plato and Aristotle, and there are references to pre-Socratic treat ments of such grammatical categories as gender.10 It is certain that Plato recognized two major components of the sentence (l6gos): 6noma and rhlma.ll At this period in the history of linguistic thought and before later developments had taken place itwould be anachronistic and misleading to translate these terms as 'noun' and 'verb'. Whether the two components were regarded as chain-exhausting by Plato cannot be made definite, but one notices that their first identificationwas made as sentence parts, parts of logos (mere logou),not as classes of words. In previous non-technical usage onoma was 'name', and rhema, as well asmeaning 'word', meant 'saying', 'proverb',and was used to refer to short Plato thus seems to bifurcate independent and often 'elliptical' sentences.12 the sentence, 16gos, into topic and comment, and would probably have assigned two stretches in it to onoma and rhema respectively, theNP and
VP of the first rewrite rhema, VP. rule of a generative grammar: l6gos, S -> 6noma, NP +

Certainly this is consistentwith much of what he says elsewhere, and later grammarians recognized the special and fundamental significance of the
binary 6noma-rhema division in sentence structure, as the basis of the two

word favourite sentence type and the two heads of furtherexpansions (nodes of subsequent branching in phrase structure).13 Semantically Plato observed the broad correlation of onoma with the
actor and rhema with the action in many sentences of this pattern.14

There has been considerable discussion on the inclusion by Plato, and later by Aristotle, of adjectives and adjectival phrases like leukos (white) and Dii philos (beloved of Zeus) among the rhemata.By the exercise of historical hindsight it is not difficult to accuse them of inconsistency and the failure to recognize the 'properly relevant criteria'. In another passage leon (lion), elaphos (deer), and hippos (horse) are given as typical onomata,
10Aristotle, Rhetorica 3.5; De sophisticis elenchis 14. 1 'roVTcov [sc. 6vopadT0v cKat5iArlsd 6q ;ydiaxt TI ov] Cratylus 425 A, 431 B: Xoyot ydp Irou E6vecis EcaTt (Sentences, I think, are the combination of these (sc. on6mata and rhemata)), Sophistes 263 D. 12 E.g. Protagoras 342 E, 343 A, B. 13Cp. E. Bach, Introduction to TransformationalGrammars,New York, 1964, 34; Apollo nius Dyscolus, De constructione 1.3 (ed. Bekker, Berlin, 1817, 22): Tx bn6Xotnra T&v
Pepd6v TOO X6you V6yECTati 7rp6o Tflv TOO 15fmaT?o Kai ToO 6v60aTo( o v TaStv (The

remaining parts of speech are referred to the syntax of the verb and the noun); Priscian, Inst. gram. 2.4.15.
14 Sophistes 262 A: T6 AIEv &'ci Taiq tpdCsolv 6v pa 8iL5o 5q)Ad otIou Xt'kyosEv ..., To

6e y' rtC' acuToi 6K}eivaip6TTOUot a cnraieov Tqnq cWvfi 6vopac (The representation IFTe9 (Tv of actions we call rhema, ... but the vocal sign assigned to those performing the actions we call'6noma). 7

This content downloaded from 198.168.27.222 on Fri, 20 Sep 2013 08:31:26 AM All use subject to JSTOR Terms and Conditions

R. H. ROBINS

and badidzei (walks), trechei(runs),and kathezdei (sleeps)as typicalrhemata15. But the first division of the Greek sentence into components had been in termsof syntactic function, not of morphological categories, and the specific identification of word classes as such had yet to be made; and in Greek adjectives and adjectival phrases served often as predicates without a part of einai (the copula verb 'to be') necessarily present; that is, they could stand alone in the rhemaplace in the structureof the sentence.Beyond this, at the stage represented by Plato, one need not, and perhaps should not, press thematter further. In particular it is to be remembered that one is not dealing herewith questions of the use or misuse of an existing technical metalanguage, but with a very early period in theworking out of just such a metalanguage from thematerial provided by Greek and not hitherto put to such a purpose. The 'relevant criteria'were only established at a later stage, as the development of Greek grammar proceeded. It may be held that grammatical theory was developed by Aristotle beyond the point attained by Plato; but we still have to abstract his specifi cally grammatical observations from various places in treatises not them selves primarily concerned with grammatical exposition. Bearing this in mind and also the fact that to look for anything like a distinct discipline of grammarwithin philosophia before the Stoics is anachronistic, one need not be too much put out by apparent or actual inconsistencies between different passages in separate works, even apart from considerations of textual genuineness in some cases. Certainly Aristotle faced the problem of formal word unity and of the selection of defining criteria for identification of different types of word from the three sources available in linguistics: word form, syntactic function, and (apparent) classmeaning.16 In Aristotle we find the first explicit treat ment of grammatically relevantword form variations later to be collected
and ordered in the familiar paradigms. At this stage ptbsis was used in

differentlyof any suchvariation inword form, on themodel of a descriptively basic shape and pt4seis derived therefrom.Aristotle thus distinguished the oblique cases and non-present tenses, as pt6seis onomatos and pt6seis rhematos, from the nominative forms, onomata, and present tense forms,
15 Aristotle, De interpretatione 1, 10; Plato, Cratylus 399 B; Sophistes 262 B. 16De auviKCrlV fivsu interpretatione 2: 6vogla Iv oSv CaTt <povfq crlmavTixi KcaTd&
Xp6vou, fq pirlG8v jRpoq atri "arlmavTc6 v cKXoptaoivov (An 6noma is a vocal sound

having a meaning by convention, not involving time reference, such that no part of it taken separately has a meaning); ibid. 3: bf5j4a 6e Iaut r6 npoaaorlatvov xpovov o0 KaO' ETrpou %syogvov crnletov (A tupoq o)68v rlatvevt X)pi?, Kai e'aTv d&i T&Ov rhema is that which in addition indicates time, but of which no part taken separately has any meaning, and it is always a sign of what is predicated of something else). 8

This content downloaded from 198.168.27.222 on Fri, 20 Sep 2013 08:31:26 AM All use subject to JSTOR Terms and Conditions

THE

DEVELOPMENT

OF

THE WORD

CLASS

SYSTEM

rhemata;comparative and superlative forms of adjectives, and deadjectival adverbs ending in -os are also referred to as ptoseis.17 That Aristotle, like Plato, made onomata and rhdmata themain constit uents of l6gos, the sentence, specifically l6gos apophantikos (declarative sentence), there is no doubt, and the Latin grammarians ascribed this binary system to him.18But he went further than Plato is known to have done in explicitly defining each. Both onomata and rhemataare semantically indivisible, and rhemata are distinguished from onomata by their function as predicates and their inclusionwithin themselvesof time reference (tense), often relevant to the truth or falsity of a proposition.Morphological criteria (tense inflections) are thus brought into the defining criteria of words for the first time, and this has been felt tomake the inclusion of adjectives such more troublesome as leukos (white) and dikaios (just) among the rhemata thanwas this same practice inPlato, mentioned above.19One may, however, note the statement by Aristotle that rhcmataby themselves,when not form ing part of a sentence, are onomata (i.e. suchword forms, like any isolated word forms, can be hypostatized, as in citation, and treated as nouns).20 Moreover, elsewhereAristotle equates single verb predicates like badidzei (walks)with copulative phrases like badidzonesti (iswalking)21; rhema,then, refers to certain sets of words functioning in their capacity as the second component of a two part sentence;mostly rhSmataare what were later distinguished as verbs, but when the later adjectives (leukos etc.) occur as predicates in a sentence like ho dnthroposdikaios (theman is just), some part of the copulative verb einai, such as esti (is) could always be addedwithout change of grammatical structureor meaning, and such a word carries time reference within its form (prossemaineikhronon). In transformational terms these verbless sentences can be generated by the deletion of some part of
einai.

In sum, the major membership of the sets of Greek words regularly serving as predicates carries time referenceas an obligatorymorphosemantic feature; but in this binary classification,without further subdivisions, one
17De 6 6voPaTa aUXa 8 interpretatione 2: T6 6( icoivo4q il [iovt Kat 6oa zotatCa obK TCT)6baet v6g1aTOq (Philonos ('of Philon') or Philoni ('to Philon') and similar forms are not onomata but pt6seis of an onoma); ibid. 3: T6Oyiavev i TO6 &a& TCTaiotS 6ytavT ob nriPa 5iflaToq (hygidnen ('he was healthy') or hygiane? ('he will be healthy') is not a rhema but a ptosis of a rhema); Topica 5.7, 2.9, 1.15.
18 Rhetorica 3.2: 6vTwv 8' 6voaTGOdV Kai rflaTCOV 9t d)V 6 k6yoq auvcalTnKEV (ondmata

and rhemata being the constituents of the sentence); Varro, De linguaLatina 8.11; Cledo nius, Ars secunda (ed.H. Keil, Grammatici Latini 5, Leipzig, 1923, 34). 19 See note 15.
20 De interpretatione 3: af)Ta V gv ov Kac' tatra Xeyo6eva Ta p5iaa 6v6oaTa aOTtV

(Uttered by themselves rhemata function as onomata). 21 De interpretatione 12;Metaphysica 5.7 (1017 a 26). 9

This content downloaded from 198.168.27.222 on Fri, 20 Sep 2013 08:31:26 AM All use subject to JSTOR Terms and Conditions

R. H. ROBINS

encounters the conflict between syntactic function and morphological form when these are both used as criterial in defining certain elements but do not in all cases converge. Severalwriters attribute a furtherdevelopment of the sentencecomponent (meros l6gou) system to Aristotle, the recognition of a third class of sjndesmoi.22 It is possible that Aristotle made some distinction in rank
between on6mata and rhemata on the one hand and the words the assignment of adverbs he collectively onomatos

designated syndesmoion the other.23The wide coverage of this third class,


together with in -5s to the ptoseis

suggests that the tripartite classification was meant to be exhaustive of the components of sentences. syndesmos includedwhat were later classed as conjunctions (syndesmoi), articles (arthra), and pronouns (antonymiai), and it is, in consequence,misleading to translate theAristotelian syndesmos as 'conjunction' without further explanation.24 At this stage in the evolution of descriptive grammar, one need devote little wonder atAristotle's alleged 'failure'to distinguishbetween thedifferent types of syndesmos.This removes the force of Lersch's argument that a
thinker separate of the status class of Aristotle must have recognized in its own right.25 Such further specification the article was as a of the work

laterwriters developing Aristotle's distinction between the classes of words essential to an independent and complete declarative sentence and those that serve subordinate functions in expansions of the basic type. Certainly this last statement does not well apply to the personal pronouns, ego (I),
for onomata sY (you singular), etc.; but these words, used as substitutes in the binary declarative sentence, are much less frequent in such a use than are nouns, since their occurrence is not obligatory in Greek but only serves sentence purposes of emphasis, contrast or the like, and so a pronoun-verb

is not precisely the grammatical equivalent of a noun-verb or 6noma-rhema


sentence, such as had been taken as the basic syntactic structure.

22 Dionysius of Halicarnassus, De compositione verborum 2, De viDemosthenis 48, and Quintilian, Institutio oratoris 1.4 couple his pupil and successor Theodectes in this. 23 Priscian, Inst. gram. 11.2.6: Quibusdam philosophis placuit nomen et verbum solas esse partes orationis, cetera vero adminicula vel iuncturas earum (Some philosophers have preferred to say that the noun and the verb are the only parts of speech, the rest being there to support them and connect them). 24Rhetorica 3.5; ibid 3.12 gives a general definition of the syntactic and semantic function of all syndesmoi: 6 yapCTUv6eSeaCo Tt ioXXha(The syndesmosmakes a unity of iv COIto themany (sentence elements)). 25 Lersch 2, 16-17. The probably spurious Rhetorica ad Alexandrum 25 distinguishes syndesmos and arthron, and the corrupt Chapter 20 of the Poetica lists drthron in a set of eight very heterogeneous mere (lekseos) (parts of discourse). Steinthal, Geschichte 1, 264 well refers to Analytica priora 1.40, where the distinction made depends on the use of theGreek article, but no mention ismade of drthron. 10

This content downloaded from 198.168.27.222 on Fri, 20 Sep 2013 08:31:26 AM All use subject to JSTOR Terms and Conditions

THE

DEVELOPMENT

OF THE WORD

CLASS

SYSTEM

The separate recognitionof the drthronto includeall the inflected members of Aristotle's sjndesmos class (i.e. the later article and pronoun classes)was specifically the work of the Stoics; and their separation of it as a distinct meros logou (or stoicheion (element), as they referred to word classes26) is part of a wider reorganization of grammatical class defining categories undertaken by the Stoics. Only after this is it legitimate to assert that the European grammatical tradition was definitely set on the lines it was to follow to the present day. From the Stoic period on one is dealing with the establishment and definition of differentword classes, each having several syntactic functions, rather than with syntactic components themselves27;
and the meaning of meros l6gou comes nearer to that of the modern 'part

of speech'. The Stoics, whose philosophical attitude led them to give linguistic science, and specifically grammar, a place of its own in their system, restricted ptosis to its subsequent limits, referring exclusively to the inflectional
differences of nouns and of words inflected in comparable paradigms. From this point it is possible to translate ptosis by the Latin casus and the English case. The Stoic drthron (covering the later article and pronoun) was case

inflected,ptotik6n; the syndesmoi that remained fromAristotle's slndesmos class were all uninflected, dptota.The Stoics further extendedptosis to cover the nominative forms of case-inflectedwords, and distinguished the ptosis orthe (or eutheia), the 'upright'case, from the pt6seis pldgiai, the oblique cases. This division was confirmed by their distinction between rhemata constructing minimally with a nominative case (e.g. Sokrdtes peripatei, Socrates walks) and those constructingwith an oblique case (e.g. Sokratei
Socrates metamelei, were mainly pldgiai Dion) regrets).28 concerned It was that the ptoseis further recognized of the verbal predicate in the extension and construction

(VP - V + NPobl.), as in sentences like Pldton Diona philei (Plato loves


or Pldton akouei Sokrdtous (Plato listens to Socrates);

26 Diogenes Laertius, Vitae philosophorum 7.192-3; Lersch, Sprachphilosophie 2, 26-27. 27 Cp. H. Koller, 'DieAnfange der griechischen Grammatik', Glotta 37 (1958) 5-40. Koller (op.cit. 28-9) regardsAristotle's svndesmos as the copula only, the third sentence component, uniting subject with predicate (6nomawith rhema).This would fit the defini tion given in Rhetorica 3.12 (note 24, above) quite well, but (1) it cannot be renconciled with the illustration inRhetorica 3.5, (2) it assumes that all the other passages exemplifying the syndesmoi in the text of Aristotle as we have it are not only corrupt but thoroughly unaristotelian in doctrine, and (3) it implies, as Koller says, that Dionysius of Halicar nassus misunderstood the position when he speaks of the Stoics 'separating out the drthra from the syndesmoi (XcopioavTcq and6 rTCv 6pOpa, De compositione ouv6o!taov Txa verborum 2). 28Diogenes, Vitae 7.43, 55-56; Steinthal, Geschichte 1, 305-6; Apollonius, De con structione 3.32. 11

This content downloaded from 198.168.27.222 on Fri, 20 Sep 2013 08:31:26 AM All use subject to JSTOR Terms and Conditions

R. H. ROBINS

with the oblique cases provided syntactic criteria for the Stoic definition of active (transitive)verbs.29 The position of the vocative case in Stoic grammar is uncertain, and among the early Stoics itmay have been excluded, their fifth case beside the nominative and the three oblique cases being derived adverbs ending in -os,which Aristotle had treated as ptoseis ondmatos.30 The first Stoic system was fourfold, the two major classes 6noma and rhema being distinguished as ptotikon and apt6ton respectively.Apt6ton is its opposite is implied, though not part of the Stoic definition of the rhMma; made explicit, in their definition of the dnoma.The Aristotelian syndesmoi were also divided intopt6tikd and dptota, the former, drthra, being further specified as gender-marking and number-marking, and including the later personal pronouns, article, and interrogatives31,and the latter, syndesmoi being designated as syntactic connectors.32 Though the Stoics made a considerable advance in the theory of Greek tenses, and distinguishedwithin the semantics of Greek tense forms the two factors of time reference and completion-continuity33, they did not apparentlymake tense part of the definition of rhemata. At a later stage,Diogenes and Chrysippus divided onoma into two classes, onoma (propernoun) and prosegoria (common noun).34The distinction was based on the alleged semantic distinction between individual quality (idia poiotes) and common quality (koine poiotes), though thiswas implausibly linkedwith a morphological criterion of different inflection,Pdris, Pdridos The formal criterion obviously will as against mantis, mdntios (prophet).35
29 c v o5v goai Tt oauv'Taa6oeva Diogenes, Vitae 7.64: 6p8a (sc. KaTrlyoplPaxTa) ;;atyiov nTdxGcoV... olov &Kcoft, 6pa ita3kyeTat (Active (verbs) are those that lti T6&v are constructed with one of the oblique cases, for example akouei (he hears), horai (he sees), dialegetai (he talks with)). 30 So Steinthal, Geschichte 1, 302; L. Hjelmslev, La categorie des cas, Aarhus, 1935, 4; contra Pohlenz, 'Begriindung', 169. 31Diogenes, Vitae 7.58: iplVpov56'<oti aTOtCeov O6youTwTOtlI6v, 8iO6piov a yvrl Kai TOUq T6)V6voLnxTGv &pt16s6uS (The drthron is an element of speech with case in flection and distinguishing the gender and number of nouns); Priscian, Inst. gram. 2.4.16, 11.1.1. 32 Diogenes, Vitae 7.58. 33 Bekker, An. Graec. 2, 891; Steinthal, Geschichte 1, 307-17. 34Diogenes, Vitae 7.57. That this was not a mere subclassification within a single class (the later position) is attested by the scholiast inAn. Graec. 2, 842: of 6 xrociKoi6v6MaTa 6 iporlyopuKa obK 6v6aTa (The Stoics called proper nouns pEv Td K6pta zXeyov,Txa onomata, and common nouns they did not call on6mata). This is the Stoic system cited by Priscian, Inst. gram. 2.4.16: Secundum Stoicos vero quinque sunt eius (sc. orationis) partes: nomen, appellatio, verbum, pronomen sive articulum, coniunctio (According to the Stoics there are five parts of speech: noun, appellation (common noun), verb, pronoun or article, conjunction). 35Diogenes, Vitae 7.58; Bekker, An. Graec. 2, 842. 12

This content downloaded from 198.168.27.222 on Fri, 20 Sep 2013 08:31:26 AM All use subject to JSTOR Terms and Conditions

THE DEVELOPMENT

OF THE WORD CLASS SYSTEM

not work; one need only think of phrontis, phrontidos (thought), but, once again, one is here seeing the process of trying out different possibilities that might suggest themselves. In a furthermodification of the Stoic system,Antipater is said to have made a six class system by the recognition of mes6tes as a word class.36 This class probably contained as itsmembers the adverbs in -os; thiswas the subsequentusage of the term,after epirrhema had become thedesignation the of class of all adverbs.37The name mes6tes may be taken from the neutralization of masculine, feminine, and neuter gender in the adjective from which it is derived, or from themiddle ground it occupies in Greek grammar in being morphologically formed from members of the 6noma class but having its principal syntactic function as part of the endocentric expansion of the rhemaclass. If previously such adverbs had been regarded as a ptOsisonomatosby the Stoics, presumably the vocative case formswere now admitted as one of the five nominal cases, where they subsequently remained. Other Stoics referred to adverbs by the term pandektes; the technical etymology of this is obscure38, but conceivably itmight be the result of widening the class to include all words syntactically equivalent to adverbs in -os (pandektes, the 'all-receiver'),themembership of the subsequent class of epirrhemata. The next documented stage in the development of the Greek word class system is the one probably established by Aristarchus, but set out in the short Greek grammar of Dionysius Thrax.39This systemwas the basis of the syntacticworks by Apollonius Dyscolus, and was passed on by him to Priscian (with the omission of the drthron,not represented inLatin, and the separate recognition of the interjection). During the final stage in the evolution of the Greek system, Alexandrian literary scholarship was the dominant context in which grammatical researchwas undertaken, and the study of classical literaturewas the channel whereby grammatical theory continued in theLatin speakingworld andwas transmitted throughPriscian and Donatus to the earlymiddle ages. The finalGreek system comprised eightword classes,with the recognition of three more distinct classes, the pronoun (antonymia), the participle (metoche), and the preposition (prothesis), and the merging of the Stoic
36 Diogenes, Vitae 7.57: 6 6 AvxinaTpoq Kai xirv pEa6crTnTa Tiqiotv vTrotiq nipi xecgcoq

Kai -&v keyotwvcov (Antipater also puts the mes6tes among the subjects treated in his Speech andMeaning). 37Dionysius Thrax, Techne, Section 24. 38 Lersch, Sprachphilosophie 2, 45-6. 39 See further Steinthal, Geschichte 2, 189-327; Robins, 'Dionysius Thrax and the Western Grammatical Tradition', TPS 1957, 67-106. 13

This content downloaded from 198.168.27.222 on Fri, 20 Sep 2013 08:31:26 AM All use subject to JSTOR Terms and Conditions

R. H. ROBINS

distinction between onomata and prosegoriai back into the single onoma class. Itmay be that the pronoun was first separated as a class by itself by theHomeric scholarZenodotus40; but one may consider the three additions to the system together. All three additional classes arise from the splittingof a previously unitary class, though in two cases a subdivision within the earlier class ismade into a definite division between two new classes, each separatelydefined. In the Stoic drthron class, 'definite'drthra horismena (personal, reflexive, and possessive pronouns) were already distinguished from 'indefinite' drthra aoristode (articles, interrogative and relative pronouns).41 The definition given to the Stoic drthra as including the distinction of genders42 is, in fact, better adapted to the words comprised by the aoristodes subdivision than to the Stoic class as a whole. In theAristarchan system personal concord with verbswas made the distinctive criterion of the pronouns, and the inter rogative pronouns were reallocated as a subdivisionof the onoma class.43 The article and relative pronouns remained within the drthron class as drthraprotassomena ('articlespreceding their nouns') and drthra hypotasso mena ('articles following their nouns'), respectively. Within the Stoic syndesmoi the laterprotheseis, prepositions, were rec ognized as prothetikoi syndesmoi,preposed conjunctions, one of a number of subclasses, the others of which remainedas subclasses in theAristarchan system.44 While we appear to have no direct references to participial forms in quotations from Stoic writers, Priscian tells us that they classed them with the verbs, as participiale verbumvel casuale, presumably rhema metochikon or rhemaptotikon, an exception to their definition of rhemaas dptoton (not inflected for case) rather awkwardly annexed to it.45 The definitions of the eightAristarchan classes, as set out by Thrax, are worth quoting, as his was taken as the final statement in antiquity of the
mere
noun.

16gou46:

onoma: a part of speech inflected for case, signifying a person or a thing;

paid particular attention to the pronouns in his Homeric textual criticism, and Apollonius Dyscolus, De constructione 2.22 refers to 'thewritings of Zenodotus on &xcovultKuc&ypa(pk; to Zrlvo86Oou). pronouns' (T dv 41Apollonius, De pronomine 4 B; Priscian, Inst. gram. 2.4.16, 11.1.1. 42Diogenes, Vitae 7.58. 43Apollonius, De pronomine 1C. 44 Apollonius, De coniunctione (Bekker,An. Graec. 2, 480); De constructione4.1; Dionysius Thrax, Techne, Section 25. 45 Priscian, Inst. gram. 2.14.6. 46Greek text in Steinthal, Geschichte 2, 210-1. 14

40 Zenodotus

This content downloaded from 198.168.27.222 on Fri, 20 Sep 2013 08:31:26 AM All use subject to JSTOR Terms and Conditions

THE

DEVELOPMENT

OF THE WORD

CLASS

SYSTEM

rhema: a part of speechwithout case inflection, but inflected for tense, person, and number, signifying an action performed or undergone; verb. metoche: a part of speech sharing the features of rhema and 6noma; participle. drthron: a part of speech inflected for case, preposed or postposed to ondmata;article. antonymia:a part of speech servingas substitute for an 6noma,andmarked
as to person; pronoun.

prothesis: a part of speech placed before other words in compounding and in syntax; preposition. epirrhema: a part of speechwithout inflection,modifying or added to a rhema: adverb. sjndesmos: a part of speech binding together the discourse and filling gaps in its interpretation;conjunction. One sees the basic categorial distinction of case-inflection and non-case inflection used to define the two fundamental inflected classes, dnoma and rhema. Given this, the separate recognition of the participle (metoche), though itsuniversal derivative statuswas laternoted47,was logically entailed. One can also trace an awareness of the particular noun-like grammatical features of infinitival forms in Greek (aparemphatonrhema, indeterminate verb, not formally specifyingperson and number). This can be seen in Stoic theory, inwhich infinitiveswere distinguished from all finite verb forms48,
to a and some grammarians are said to have regarded them as belonging was not of the choice de class.49 determined This, however, separate by

fining criteria as was the status of the participle, since absence of case inflectionhad been made the principal definiens of the verb, and infinitives do not exhibit overt case markers, though inGreek they are constructible with case-inflecting articles.50 The distinction between inflectedwords (lekseis klitikai) and uninflected or invariablewords (lekseis dklitoi or ametakinetoi) is the differentiating

47 Cp. scholiast, An. Graec. 2, 896: aei Ev nTapayoy, esaTiV.((The participle) is always a derived form); Priscian, Inst. gram. 11.1.2: Semper in derivatione est (It is always a derived form). 48Apollonius, De constructione 1.8: ot dic6 tfq ZTodqsabr6 (To dicapeuqcparov) tev IKaXoicnl p5ga, TO8f cEptRarctt KarTiy6prlga r aOPj'laga. (The Stoics call the I ypda(pet infinitive itself the verb, and forms like peripatei (hewalks) or grdphei (he writes) they call predicates). 49 Priscian, Inst. gram. 2.4.17. 50 Scholiast quoted by Steinthal, Geschichte 2,287: xla sesa (dppou Xay6geva&caap EqpaTa 6v6oagaa gdb6v cai fv qi grata (Infinitives preceded by an article are nouns rather than verbs). 15

This content downloaded from 198.168.27.222 on Fri, 20 Sep 2013 08:31:26 AM All use subject to JSTOR Terms and Conditions

R. H. ROBINS

factor separating the first five classes from the last three.Thrax only makes this explicit in his definition of the adverb (epirrhema),but the scholiasts add it to their definitions of the other two.51 In his exposition Thraxmakes quite explicit the differencebetween distinct classes, each with a separate definition covering the entiremembership of the class, and various subdivisions or subclasseswithin it. The Stoic onoma (proper noun) and prosegoria (common noun) are regrouped as subclasses within the onoma class.52 The later adjectives are similarly treated as a subclass of on6mata, onomata epitheta, nomina adiectiva, whence the use, still sometimes found in modern linguistic literature of the terms noun substantiveand nounadjective.One observes that themorphological criterion for distinguishingGreek (and Latin) adjectives, the existence of a paradigm of gradation (positive, comparative, and superlative forms), was excluded in the system set out by Thrax, by his allocation of such formsas andreioteros, braver, and oksftatos, swiftest, to separate subclasses of the onoma, along with others.53 The class of particle, recognized in some grammars of Greek to-day, though of indeterminatemembership, falls within various subclasses of Thrax's syndesmoi(e.g. symplektikoi, linking,men, de, etc., parapleromatikoi, expletive, dr, an, ge, etc.).54 To speak of the system set down by Dionysius Thrax as final is not to
imply that in the ensuing and tradition of no further changes the classes was at all were made. cases altered.

Some of his definitions were criticized and some different ones were put
forward55, the membership in some

The Latin recognition of the interjectionclass and the necessary suppression of the article class have been mentioned. Morphological similarities had enabled the Greeks to group their relative pronoun h6s, he, ho with the definite article ho, he, to, as drthra hypotassomenaand drthraprotass6mena respectively.The Latin equivalent to the drthonhypotass6menon,qui, quae, quod, along with itsmorphologically similar partner quis, quae, quid, was
allocated either to the nomen class or to be pronomen class, in which latter

it remains to-day.56
51 Bekker, An. Graec. 2, 924, 952. 52 Dionysius Thrax, Techne, Section 13; Priscian, Inst. gram. 2.5.22-24. 53Dionysius Thrax, Techne, Section 14: 6vopa ruyKpttlc6v and 6vola 6nRcpcriTKOV; Priscian, Inst. gram. 3.1.1, 3.3.18: Nomen comparativum, nomen superlativum. 54Cp. J. D. Denniston, The Greek Particles, Oxford, 1954. 55 E.g. Scholiast, Bekker, An. Graec. 2, 952 (on the conjunction): oUVKctK0c6v TxV c auaorllTdtvst (Binding together the other parts of speech to oto Xo6you ptsp&v o1 Kai whose meaning it also contributes); so Priscian, Inst. gram. 16.1.1. 56Greek tiq (who?) was an 6noma ertetmatik6n (interrogative noun) inDionysius Thrax, Techne Section 14; see too Apollonius, De pronomine 33-5, where the assignment of tiq to the onoma class is discussed and supported. 16

This content downloaded from 198.168.27.222 on Fri, 20 Sep 2013 08:31:26 AM All use subject to JSTOR Terms and Conditions

THE

DEVELOPMENT

OF THE WORD

CLASS

SYSTEM

But the descriptive frameworkof eight classes remained,and thegrammat ical categories thatwere taken as criterialwere themselvesunaltered, though theywere differently applied in some details. Indeed, somuch was Priscian's treatmentof word classes in Latin accepted, thatwhen in the latermiddle ages the speculative grammarians placed quite a different interpretationon Latin grammar, requiring among other things explanatory adequacy as against mere descriptive accuracy, it was never thought necessary to call
in question the classes and categories that Priscian had used to make his

Nor do modern presentations of Greek and Latin grammatical system.57 grammar, though differing in their treatmentof the adjective as a separate word class and of the participle as part of the inflectionof the verb, depart radically from the systematic framework set down by Dionysius Thrax and
Priscian.58

Throughout the course of its elaboration, and in its final form, ancient grammar relied on definitions of its classes in a basicallyAristotelian form, in terms of generic and specific features, categories, and attributes. The definition of the word class adjective in Hill's Introduction,though framed for English alone and couched inmodern morpheme distribution termino logy, is of the same type as the definitions given by Thrax: 'Anyword with
of slow and capable of being modified by the addition of -er -est is an adjective'.59 These definitions are in a form that allows for to be allocated to a given class on words newly invented or encountered the basis of its explicit definition (hard or marginal cases may arise, as they the distribution and have always arisen in any classificational categories system, but that is not or affixes are to be made the point criterial

here). Such definitions are not discovery procedure statements, in that one
is not told why any particular Latin qui and quis as part of the nomen class, Priscian, Inst. gram. 2.4.16, 2.6.30, 13.3.11, 13.4.21; as part of the pronomen class, Probus, Institutio artium (ed.Keil 4, 133),Donatus, Ars grammatica (ed. Keil 4, 379). 57 Cp. the complaint of William of Conches: 'Quoniam in omni doctrina grammatica precedit, de ea dicere proposuimus, quoniam, etsi Priscianus inde satis dicat, tamen obscuras dat definitiones nec exponit, causas vero inventionis diversarum partium et diversorum accidentium in unaquaque praetermittit.' (Since in all learning grammar takes precedence, we have set ourselves to deal with it, because, though Priscian states it adequately, his definitions are obscure and he does not explain them, and he passes over the causes of his setting up the various parts of speech and the various accidents to which each is subject) (H. Roos, 'Die Modi Significandi des Martinus de Dacia', Beitrdge zur Geschichte der Philosophie und Theologie des Mittelalters, 37.2 (1952). 93). 58Cp. C. D. Buck, Comparative Grammar of Greek and Latin, Chicago, 1933, 168; B. H. Kennedy, Revised Latin Primer, London, 1930, 12; R. Ktihner, Ausfiihrliche Grammatik der griechischen Sprache 1,Hanover, 1890, 355-6; id.Ausfiihrliche Grammatik der lateini schen Sprache, Hanover 1912, 253-4. In these two last books, the separate chapters on the numerals (621 and 629, respectively) aremerely a pedagogical rearrangement of certain members of the noun, adjective, and adverb classes. 59A. A. Hill, Introduction to Linguistic Structures, New York, 1958, 168. 17

This content downloaded from 198.168.27.222 on Fri, 20 Sep 2013 08:31:26 AM All use subject to JSTOR Terms and Conditions

16gos
00oo Plato 6noma m

Aristotle

6noma

rhema

Stoics

6noma

rhema

Stoics

6noma

prosegorfa

rhema

Stoics

6noma

prosegoria prosegonia

mes6tes (pand6ktes)

rhma rh~iema s

Dionysius (Aristarchus)

onoma

epfrrhema

rhema

metochC

prothes

Priscian

nomen

interiectio

adverbium

verbum

participium

preposi

adjective

noun

interjection

adverb

verb

prepos

Diagram showing the development of the word cla

This content downloaded from 198.168.27.222 on Fri, 20 Sep 2013 08:31:26 AM All use subject to JSTOR Terms and Conditions

THE

DEVELOPMENT

OF

THE

WORLD

CLASS

SYSTEM

or why a particular set of classes is themost suitable. Such work has already been done. Nothing was said expressly on fieldmethods by the ancients, though the testing and refining of rough categories and classes through successive stages (as with the Aristotelian ptbsis and syndesmos) can be seen as an extended bit of field discovery spread over several generations. Bach contrasts Hill's definition, given above, with a generative definition
in the form of a rewrite rule (Adjective -, X), whose right hand component

is a list (perhaps recursive).60Such an enumerative definition is a stage ahead, in as much as considerations of categories, distribution, and para
digms have already been taken into account in the decision to frame the

rules in thisway andwith a given listing.Neither type of definition is strictly 'operational', nor concerned with the distinction between discovery by a flash of intuition and discovery by laborious sifting of material (presumably The reasons for guided by some sort of intuition if it is to get anywhere).61 rules being framed in a given form must be recoverable, just as adequate criterial definitions can be rephrased as rules; and the reasons for a listing rulemust be such that an existing grammar of a living language can take on new vocabulary creations and changes in the grammaticalusage of words and incorporate them into its existing constitution. This is not the place for a comparative evaluation of the grammar of formal definitions and the grammar of rules.62 What one may legitimately stress is that, in tracing the genesis, development, and fixation of theword class system of classical grammar in western antiquity, we face problems and controversies very much at the forefront of contemporary linguistics,
and at the same time follow the evolution of a categorial support that has

upheld and guided some 2000 years of continuous and not inconsiderable linguistic scholarship. School of Oriental andAfrican Studies, University of London

60 Bach, Introduction, 28-9. 61 Ibid. 186.


62 Cp. W. O. Dingwall, 'Transformational Grammar: Form and Theory', Lingua 12 (1963)

233-75. 19

This content downloaded from 198.168.27.222 on Fri, 20 Sep 2013 08:31:26 AM All use subject to JSTOR Terms and Conditions