GREEK WORD ORDER BY K. J. DOVER Professor of Greek in the University of St Andrews CAMBRIDGE AT THE UNIVERSITY PRESS 1960~CONTENTS Preface Bibliography I THE NATURE OF THE PROBLEM (i) Indeterminacy (ii) Types of determinant (ii) Material Il LEXICAL AND SEMANTIC DETERMINANTS Gi) Postpositives and prepositives (ii) Types of clause (iii) Preferential words Il SYNTACTICAL DETERMINANTS (i) General principles (ii) Preferential words (iii) Ordinary words IV LOGICAL DETERMINANTS (A) General principles G@) Emphasis (ii) Logical categories (B) Concomitants (i) Treatment of concomitants as postpositives (ii) Concomitant groups (iti) Treatment of concomitants as prepositives (iv) Preferential treatment of concomitants (c) Nuclei Series (ii) Formulae Vv STYLE Index of passages Page v ix 10 I2 12 19 20 25 25 26 28 32 32 32 34 41 4 46 48 49 53 53 56BIBLIOGRAPHY The abbreviations of periodicals are those of L’ Année Philologique. Bibliographies of Greek word order are given by Schwyzer (Gram- matik), Schwenter in W & S vii (1923), pp. 179 ff., 1X (1924), pp. 194f., N.F. 1 (1938), pp. 160ff., and Maurer, ibid. 1x (1924), pp. 195f. These include references to works of the early and middle nineteenth century. Some modern work is briefly mentioned by Schwyzer in JAW cc (1924), pp. 138ff., and Risch, G/ xxxv (1956), pp. 33 ff. There is a good biblio- graphy of general linguistic works in Sandmann (see below). The articles of Firbas (see below) contain many references to work on individual modern languages, including work in and on Slavonic languages. The works listed in the following bibliography are referred to in my footnotes (i) by the author’s name only, if the author appears only once in the bibliography, (ii) by the author’s name and an abbreviation of the title, if the author is represented by more than one work. I mark * those works which seem to me to be of exceptional value to the student of Greek word order. Ammann, H. ‘Die Stellungstypen des lateinischen attributiven Adjecti- vum und ihre Bedeutung fiir die Psychologie der Wortstellung auf Grund von Ciceros Briefen an Atticus’, JF xx1x (1911/12), pp. 1 ff. *Vom doppelten Sinn’, SHAW (ph.-hist.) 1920, no. 12. * Untersuchungen zur homerischen Wortfolge und Satzstruktur, 1, Freiburg i. B., 1922. ‘Untersuchungen, etc’, 11, JF xL11 (1924), pp. 149ff., 300ff. —— ‘Wortstellung und Stilentwickelung’, G/ x11 (1923), pp. 107 ff. BacHHaMMER, F. Die Spaltung eng verbundenen Worter bei Thukydides, Diss. Erlangen, 1921. Barse.enet, D. De la phrase d verbe ‘étre’ dans Pionien d’Hérodote, Paris, 1913. “Sur la place de yivoyen dans la phrase d’Hérodote’, MSZ xix (1914), pp. 105 ff. Bantu, P. ‘Zur Psychologie der gebundenen und der freien Wortstel- lung’, Philosophische Studien xxix (=Festschrift Wilhelm Wunde), pp. 22 ff. Bewacuet, O. ‘Beziehungen zwischen Umfang und Reihefolge von Satzgliedern’, JF xxv (1909), pp. 110 ff. — ‘Zur Stellung des Verbs im Germanischen und Indogermanischen’, ZVS 1.1 (1929), pp. 276 ff. Bercaicne, A. ‘Essai sur la construction grammaticale’, MSL m1 (18- 78), pp- 1 ff, 124ff., 169 ff. [Left incomplete by author’s death.] Bisuop, C. E. ‘The Greek Verbal in -TEO’, m, AJPA xx (1899), pp. 241 ff. *x BIBLIOGRAPHY Buocu, A. ‘Uber die Entwicklung der Ausdrucksfihigkeit in den Sprachen des Altertums’, MH 1 (1944), pp. 234 ff. BLoomFIELp, M. ‘On the variable position of the verb in oldest Sanskrit’, IF xxx1 (1912/13), pp. 156 ff. Boupt, H. De liberiore linguae graecae et latinae collocatione verborum capita selecta, Diss. Géttingen, 1884. *Botincer, D, L. ‘Linear Modification’, PMZA ixvu (1952), pp. 1117 ff. Brucmann, K. Kurze vergleichende Grammatik der indogermanischen Sprachen, Leipzig, 1904. “Verschiedenheiten der Satzgestaltung nach Massgabe der seelischen Grundfunktionen in den indogermanischen Sprachen’, BSG (ph.- hist.) Lxx (1918), no. 6. Cuantratng, P. ‘Les recherches sur l’ordre des mots en grec’, AFC v (2952), pp. 71. Detsriics, B. Die altindische Wortfolge (= Syntaktische Forschungen 11), Halle, 1878. Vergleichende Syntax der indogermanischen Sprachen 11, Strassburg, 1900. *Germanische Syntax 1: Zur Stellung des Verbums’, 4SG (ph.- hist.) xxvim (1911), Nr. 7. Denniston, J. D. Greek Prose Style, Oxford, 1952. —— The Greek Particles, 2nd ed., Oxford, 1954. Dirrmar, A. Syntaktische Grundfragen, Grimma, 1911. Exseinc, H. L. ‘Some Statistics on the Order of Words in Greek’, Studies in Honor of Basil L. Gildersleeve (Baltimore, 1902), pp. 229 ff. Entwist.e, W. J. Aspects of Language, London, 1953. Frraas, J. ‘Thoughts on the Communicative Function of the Verb in English, German and Czech’, Bmo Studies in English 1 (1959), pp. 39 ff. * ‘Some Thoughts on the Function of Word Order in Old English and Modern English’, Zvidini Otisk x Casopisu, Sbornik Praci Filos. Fak. Brn. Univ., 19§7, pp. 72. —— ‘On the problem of Non-Thematic Subjects in Contemporary English’, ZvléStnf otisk: Casopis pro moderni filologii, xxx1x, pp. 22ff., 166ff. [In Czech, with English summary.] Fiscuer, P. ‘Zur Stellung des Verbums im Griechischen’, G/ x11 (1924), pp. 1ff., 189 ff. FRaENKEL, Epuarp. Jktus und Accent im lateinischen Sprechvers, Berlin, 1928. * ‘Kolon und Satz: Beobachtungen zur Gliederung des antiken Satzes’, NGG (ph.-hist.) 1932, pp. 197ff., 1933, pp. 319 ff. *Frisk, H. Studien zur griechischen Wortstellung, Géteborg, 1933. VON DER GaBELENTZ, G. ‘Ideen zu einer vergleichenden Syntax’, ZVPs V1 (1869), pp. 376.BIBLIOGRAPHY xi VON DER GABELENTZ, G. ‘Weiteres zur vergleichenden Syntax’, ZV-Ps vimt (1875), pp. 129ff.,300ff —— Die Sprachwissenschaft, 2nd ed., Leipzig, 1901. Garpiner, A. H. Speech and Language, Oxford, 1932. *Goopktt, T. D. ‘The Order of Words in Greek’, TAPAA xx1 (1890), pp. 5. Harcuer, Anna G. ‘Syntax and the Sentence’, Word xu (1956), pp. 224 ff. Haupt, Moritz. Opuscula 11, Leipzig, 1876. Havers, W. ‘Zur “Spaltung” des Genetivs im Griechischen’, JF xxx1 (1912/13), pp. 230ff. —— Handbuch der erkldrenden Syntax, Heidelberg, 1931. Hermann, E. ‘Gab es im Indogermanischen Nebensitze?’, KZ xxximt (1894), pp- 481 ff. Hirr, H. Indogermanische Grammatik vu, Heidelberg, 1937. Hotwerpa, A. E. J. Specimen literarum continens disputationem de dispositione verborum in lingua graeca, in lingua latina, et apud Plutarchum, Utrecht, 1878. Jacossonn, H. ‘Tagen baren Léneborger Kind’, ZVS tv (1927), Pp. roof. —— ‘Zum homerischen dotepov trpétepov’, ZVS ivi (1929), pp. 1 ff. Kamer, G. Stil und Text der todrtela "A@nvatav des Aristoteles, Berlin, 1893. Kent, R. G. Old Persian, 2nd ed., New Haven, 1953. *Kiecxers, E. Die Stellung des Verbs im Griechischen und in den ver- wandten Sprachen, Strassburg, 1911. —— ‘Die Stellung der Verba des Sagens in Schaltesatzeh im Griechischen und in den verwandten Sprachen’, JF xxx (1912), pp. 145 —— ‘Zur oratio recta in den indogermanischen Sprachen’, 1, JF xxxv (1915), pp. 1f.; 1, JF xxxvi (1916), pp. 1 ff. Krausz, W. ‘Die Entwicklung einer alten elliptischen Konstruktion in den indogermanischen Sprachen’, ZVS xu (1924), pp. 223 ff. -Kinner, R. Ausfithrliche Grammatik der griechischen Sprache: Zweiter Teil: Satzlehre, revised by GERTH, B., 2 vols., Hanover and Leipzig, 1898-1904. Leo, F. ‘Bemerkungen iiber Plautinische Wortstellung und Wort- gtuppen’, NGG (ph.-hist.) 1895, pp. 415 ff. Leumann, M. and Hormann, J. B. Lareinische Grammatik, Munich, 1928. *Linpuamer, Luise. Zur Wortstellung im Griechischen, Diss. Munich, 1908, *Lorpre, A. Die Wortstellung im griechischen Sprechsatz, Diss. Freiburg i, d. Schw., 1940. McDonatp, W. H. ‘Linguistic Examination of an Epigraphic Formula’, AJA 11x (1955), pp. 151 ff.xi BIBLIOGRAPHY Macponett, A. A. Vedic Grammar, Oxford, 1916. Marouzeau, J. L’ordre des mots dans la phrase latine, 3 vols., Paris, 1922~49. L’ ordre des mots en latin, Paris, 1953. Meiuet, A. Introduction d I’étude comparative des langues indo-euro- péennes, 7th ed., Paris, 1934. Muster, K. ‘Der syntaktische Gebrauch des Genetivs in den kretischen Dialektinschriften’, ZF xv111 (1905/6), pp. 133 ff. Mupen, A. W. The Limitations of the Predicative Position in Greek, Diss. Baltimore, 1900. Moiuer, R. ‘De attributo titulorum s. V Atticorum observationes quaedam’, Ph LXIV (1905), pp. 554 ff. Newrsass, R. Sprache und Stil der Iamata von Epidauros, = Ph Supplbd. XXVII (1935), NO. 4. Pantazes, M. Td Si1acagpntixoy tis fAANVIKi}s yAdoons, ’APnve x1 (1899), pp. 443 f. Paut, H. Principien der Sprachgeschichte, 2nd ed., Halle, 1886, Pirrman, R. S. ‘Nuclear Structures in Modern Linguistics’, Language XXIV (1948), pp. 2874. Porzic, W. Review of Ammann, Untersuchungen 1, IF xutv (1927), pp. 94ff. Posteate, J. P, ‘Flaws in Classical Research’, PBA 1907-8, pp. 161 ff. Rapermacuer, L. ‘Griechischer Sprachgebrauch vu—xiv’, PA Lx (1904), pp. 1 ff. *Rass, H. J. Der Satzanbau: Untersuchung ber eine eigentimliche Wortstellung im griechischen und im lateinischen, Diss. Gdttingen, 1950. [Typescript] *Ricuter, Euisz. ‘Grundlinien der Wortstellungslehre’, ZRPA x1 (1920), pp. 9 ff Roserts, W. Ruys. ‘A Point of Greek and Latin Word Order’, CR XXVI (1912), pp. 177 ff. *SanpMann, M. Subject and Predicate, Edinburgh, 1954. Scnicx, Caria. ‘La lingua delle iscrizioni’, RF N.S. xxx (1955), pp. 361 ff. Scuéne, H. ‘Verschrinkung von Redegliedern im wiedererzahlten Dialog’, RAM xiv (1899), pp. 633 ff. ‘Verschiedenes’, RAM Lxxint (1920-4), pp. 137 ff. ‘Eine umstrittene Wortstellung des griechischen’, H 1x (1925), pp. 144 ff. Scuutze, W. ‘Eine eigentiimliche Wortstellung’, ZVS uv (1927), pp. 301 ff. Scuwyzer, E. Griechische Grammatik (completed by Dzsrunnr, A.), Munich, 1940. Spice, F. Vergleichende Grammatik der altérdnischen Sprachen, Leipzig, 1882. *,BIBLIOGRAPHY xiii Sprrzer, L. ‘Nachtrag zu Singen und Sagen’, ZVS u1v (1927), p. 311. *Tuomson, G. ‘The Postponement of Interrogatives in Attic Drama’, CQ xxx (1939), pp. 147ff. Tuums, A. ‘Experimentelle Psychologie und Sprachwissenschaft’, GRMS 1m (1911), pp. 1ff., 65 ff. Vauten, J. ‘Ober zwei Briefe des Alciphron’, SPAW 1908, pp. 990ff. Venpryes, J. Le longage, revised ed., Paris, 1950. *pe Varies, J. Untersuchungen iber die Sperrung von Substantiv und Attribut in der Sprache der attischen Redner, Diss. Freiburg i. B., 1938. *WacxzrnacrL, J. ‘Uber ein Gesetz der indogermanischen Wort- stellung’, JF 1 (1892), pp. 333. (= Kleine Schriften 1, Gottingen, 1953, pp 1ff.). —— Uber einige antike Anredeformen, Gottingen, 1912. —— ‘Zur Wortfolge, besonders bei den Zahlwértern’, Festschrift Gustav Binz, Basle, 1935, pp. 33 ff. (=KZ Schr. 1, pp. 236ff.). —— ‘Eine Wortstellungsregel des Panini und Wincklers Aleph-Beth Regel’, JF ivi (1938), pp. 161 ff. (= Ki. Schr. 1, pp. 4344f.). ‘Indogermanische Dichtersprache’, Ph xcv (1942), pp. 1 ff. *Wert, H. De Pordre des mots dans les langues anciennes comparées aux langues modernes, 31d ed., Paris, 1879. Wiuetm, A. ‘Neue Beitrige zur griechischen Inschriftenkunde’, v1, SAWW cxxxxi11 (1921), no. 3. —— ‘Zur Syntax des griechischen’, NGG (ph.-hist.) N.F. ur (1938), pp. 118 ff. Wonpt, W. Vilkerpsychologie I. Die Sprache, 2 vols., Leipzig, 1900.+ CHAPTER I THE NATURE OF THE PROBLEM (i) INDETERMINACY The problem of Greek word order is so seldom discussed in this country that it is still possible to treat it as a fresh problem. In this respect it differs from all other problems of comparable magnitude in the study of the Greek language. Of the work so far published on Greek word order, much is cautious and limited in scope, confining itself to the minute analysis of a small number of particular pheno- mena. Work which aims more ambitiously at a high level of generalisation has borne little fruit. Sometimes this is because the ‘rules’ which it offers prove inadequate, or break down, as soon as one tries to apply them to a page of Greek in a text opened at ran- dom.' Sometimes the rules admit neither of proof nor of disproof, because the arguments on which they rest are circular.? More often, an objective general statement of the facts appears, to the seeker after rules, inconclusive in the extreme; it amounts to saying ‘xyz and xzy occur, but, on the other hand, yxz, yzx, zxy, and zyx also occur.’ General accounts of word order of the kind which we find in Schwyzer’s grammar would have delighted Sextus Empiricus. It may be argued that if attempts to establish general rules have ended in doubt and confusion, that is the fault not of the enquirers after the truth but of the truth itself, Greek word order being ‘free’, ‘arbitrary’ or ‘indeterminate’. I do not suggest that such a view is unscholarly or disreputable. It speaks, after all, with the voice of experience and can draw upon a superabundance of evidence. We find an Attic boundary-stone which proclaims itself (/G, 7, 877) [40]86 46[p]os; we find another, of the same date, which says (ibid. 878) Ad[pos] 4083. The beginning of chapter 8 of the Hippocratic * Thumb, p. 2, has no difficulty in disposing, by this test, of Kiihner’s general rule (11, p. 595), but the rule which he substitutes (‘ middle’ position of the verb) can be disposed of with equal ease. 2 See especially ch. rv, a (i) below (p. 32). Similar criticisms of circularity may be brought against, e.g., Ammann’sattempt (WSt) to represent differences of Greek order by stylistic differences in translation and Richter’s treatment (p. 24) of ‘Rhythmuslinien’, 1 DG2 GREEK WORD ORDER work De Carne is 1d 5t trap &8e Euvéotn; the next chapter begins 6 8 ont Euvéotn Se. Discouraging as such examples may be, we should regard them as opening the question, not as closing it. They suffice to show that there must be some degree of indeterminacy in Greek word order; that is to say, it must sometimes be determined by processes in the mind of the composer, rational or irrational, which we cannot necessarily expect to recover. On the other hand, there are very many utterances in Greek which conform to one or other of a limited number of patterns." Take, for example, a complete utterance consisting of the three words tévra av typayev. Mathematically, there are six possible ways of arranging three unlike objects; but out of the six possibilities in this utterance, the majority can be discarded: G) &v mévta typayev, (ii) av Eypaye trdvta. No competent student would write either of these; and any competent student should be able to say why, namely that &v never begins a sentence. (iti) 2ypowe Trévt’ &v. Nor, I hope, would a student write this; but I should be surprised if he could say why, except to say (truly) that if we search for an example, in Classical Greek prose, of verb -+1r&vta-+4v arranged, as a complete utterance, in this order, our search will be long. All of us, teachers and students alike, constantly avoid abnormal formulations without knowing why, and there is nothing strange in that; the process is comparable with the familiar experience of carrying out a manual operation rightly when we do not attend to it too closely and wrongly when we do. (iv) tra&vr’ Bypaev &v. This is described by Demetrius, Eloc. 256, as a Kaxdpevov deviation from tra&vta av éypayev, but some- thing which one might say or write in order to achieve Sewdtns; he mentions it with tapeyéveto obyi=ot Tapeytveto. (v) Eypowev &v travra and (vi) 1év7’ dv éyponyev. On these two alternatives it is enough for the moment to remark that it is (vi), not (v), which Demetrius treats as the norm from which wévT’ Eyponpev &v is a deviation, and (vi) which we should be more likely to write in a prose composition; reasons for this will be given later. If it is objected that an example containing a word like év is not a fair example, it is legitimate to answer: what is a word ‘like &v’, and which words are ‘like &v’? And where does the boundary lie * Cf. Vendryes, p, 167.THE NATURE OF THE PROBLEM 3 between the comparative rigidity of wavr" dv Eypowev and the total laxity of Spos S800 / S500 Spos? The purpose of this book is to discover the nature of the principles which would justify us in calling wévt’ &v fypoyev ‘normal’ or ‘right’ in order. Gi) TYPES OF DETERMINANT . The respects in which two utterances may be similar or dissimilar are clearly very numerous. If two utterances are syntactically identical, but differ in order, this does not prove that the determi- nants of their different orders are unknowable; it proves only that syntactical identity does not suffice to determine identity of order; and our task becomes the exploration of all the respects in which the two utterances are dissimilar, in the hope of finding there the vital difference which determined their difference of order. Let us now take a very simple Greek utterance, the words with which Hippocrates excitedly awakens Socrates at the beginning of Plato’s Protagoras (3108): Mpatayépas f\xer.? Let us discard any beliefs which we may previously have entertained on the determi- nation of word order, and ask in complete innocence: Why, in this particular utterance, does the word Tparroryépos precede the word fixer? We cannot know a priori what kind of answer we shall get, if we ever get one; we must be prepared for an answer in any of the following categories: Gi) In phonological terms, e.g. ‘the longer word precedes the shorter’, or ‘a word beginning with a consonant precedes a word beginning with a vowel’. (ii) In morphological terms, e.g. ‘nouns precede verbs, irrespec- tive of their syntactical interrelation’. (iii) In syntactical terms, e.g. ‘ the subject precedes the predicate’. (iv) In semantic terms, e.g. ‘words denoting motion come at the end’. (v) In lexical terms, e.g. ‘fixer is one of n words which always come at the end’. (vi) In logical terms, e.g. ‘Hippocrates knows that Socrates will guess that someone has come, but he won’t know who; so he utters * This example is discussed by Goodell, pp. 30f., and Denniston, Prose, p. 44f. ? Cf. Jacobsohn (ctr. Spitzer) and Wackernagel, W’stregel.4 GREEK WORD ORDER first the word which is least expected by his hearer’. Here and subsequently I use the word ‘logical’ in a highly general sense, as an adjective corresponding to the noun ‘thought’.? (vii) In emotive terms, e.g. ‘the word TTpctorydpas comes first because it is the focus of the speaker’s emotion’. (viii) In social or ceremonial terms, e.g. ‘Protagoras is older than the speaker’ or ‘fxe1 is a tabu word in the speaker’s family’.? (ix) In terms involving the individual history of the speaker, e.g. ‘the last time he said, or wrote, or thought “Protagoras has arrived” it was in the form Tpwtaydpas Ket, and the present example is determined by habit’. (x) In stylistic or aesthetic terms, e.g. ‘the last time he thought or said or wrote “ Protagoras has arrived”, it was in the form 7jKet TIparterydpos, and he likes variety of formulation’. T should not like to think that these ten categories are an exhaustive list;3 but they are something to go on with. The customary pro- cedure in an enquiry of this kind is: @) To select one of these categories for exploration. Gi) To formulate, within that category, hypothetical rules of the highest level of generality compatible with their being mutually exclusive. (iti) To make a selection of texts which is random from the point of view of the rules to be tested but not necessarily random from the point of view of history, dialect or genre. (iv) To compile statistics of the observance and non-observance of the rules. (v) On the basis of these statistics, to distinguish between ‘normal’ and ‘abnormal’ order. (vi) To discover the determinants of abnormality; that is to say, to discover conditions which are present in all the abnormal * Goodell, pp. 14ff., uses ‘rhetorical’ in something like this sense. * I exaggerate here, but cf. Wackernagel, Wf. pp. 47f., for respects in which word order may reflect social conditions and changes in order changes in conditions. 3 One must also consider e.g. ‘mimetic’ order, in which words denoting sudden noise or movement precede the words denoting its cause or source (Ammann, Unt. 1, p. 15, Dopp. p. 23; Havers, Hdb. p. 146), and ‘excitant’ order, in which the words which are essential to the hearer’s understanding of the whole situation communicated are postponed in order to create tension (Lindhamer, p. 71; de Vries, pp. 87ff.).THE NATURE OF THE PROBLEM 5 instances but absent from all the normal, or, if present in some normal instances, are counteracted there by certain other conditions which are absent from the normal. The process of counteraction may, of course, be complex, so that we eventually find ourselves formulating rules which constitute exceptions to an exception to an exception to an exception to an exception to the general rule. This procedure is not only customary,' but in some form or other inevitable. Yet the difficulties and dangers inherent in it arenumerous and remarkable. (i) Statistics may or may not give a clear picture; we cannot know until we have them. If, for example, we are testing the syntactical tule ‘the subject precedes the predicate’, and we find that the ratio of SP to PS is 10:1, we can embark with some confidence on the next stage of the enquiry, the discovery of the deter- minants of abnormality. On the other hand, if the ratio is 2:1 or lower, it is unlikely that we have discovered a primary determinant of order and more likely that we are on the track of a secondary phenomenon. Gi) It is easy, but wrong, to equate ‘statistically normal’ with ‘natural’ and ‘statistically abnormal’ with ‘distorted’, ‘inverted’, etc. If, for example, we were investigating a language in which the order SP was invariable in statements and the order PS invariable in questions, it would be misleading to formulate the rule in terms such as: ‘SP is normal, but this normal order is reversed in ques- tions’. Statements are more numerous than questions, but that is not a fact of a kind with which we are concerned; we should content ourselves with the pair of discoveries that (2) statement determines the order SP, (5) interrogation determines the order PS. (iii) It is also easy, but equally wrong, to define the determinants of abnormality positively and to leave those of normal order negatively defined. Suppose, for example, that there is a language in which the order PS is statistically abnormal and in all instances of PS the predicate is the focus of the speaker’s emotion. It is mis- leading, in such a case, to say that the predicate precedes when it is * The distinction between ‘habitual’, ‘traditional’, ‘usual’, ‘banal’, or ‘natural’ order on the one hand and ‘occasional’ or ‘rhetorical’ order on the other is fundamental in Brugmann, Vl. p. 677, Delbriick, Vgl. pp. 38, 110ff., Aktind. p. 13, Schwyzer, Gr. 1, p. 691, Leumann, p. 610, Kieckers, St. Vos. pp. 2ff.; cf. Chantraine, pp. 71f., Vendryes, p. 168.6 GREEK WORD ORDER emotionally emphasised but omit to say that the subject also pre- cedes when it is emotionally emphasised. It would be more useful to say that (2) the element which is the focus of emotion precedes, and (4) a syntactical determinant operates when emotion is absent. (Perhaps here, as in other aspects of linguistics, there is room for ‘zero grade’ as a positive concept.) We might also find in some language that normal and abnormal orders are equally positive manifestations of a single general principle. (iv) No scholar, I suppose, thinks that he needs to be warned against regarding the order of his native language as ‘natural’, but the danger is more complex and more insidious than is commonly realised. The concepts ‘nature’ and ‘instinct’ die hard in linguistics. Tf in our browsing in linguistic literature we encounter an illustra- tion of word order drawn from a North American Indian language, we think it fantastic; but because most of us who know Greek at all began to learn it as children, and became accustomed to it gradually, we fail to see that the fundamental differences in order between Plato’s language and ours are at least as great as that between Sitting Bull’s and ours.t One consequence of this failure is our tendency to regard as ‘natural’ such elements as are common to Greek and English order; we therefore seek explanations of the differences, but do not trouble to explain what is familiar; theses are written about Kal é&yAak S&6o1 &rrowe (//. 1, 23)? but not about gépoov 1” Strepelon’ Etrowa (ibid. 13).3 The fallacy is not wholly irrational in origin. Since Greek and English are the result of differentiation from much more obviously related originals, and since there exist language-families in which all the members of the family follow the same principles of word order, it might be sup- posed that we can find a Greek norm of order which will also be an Indo-European order and will be manifested also in English, French and German. We might find this; but we have no right whatsoever to expect that we shall find it. Indeed, when we reflect that the ™ Cf. especially the phenomena discussed by Vahlen, p. 1002, Schéne, Umstr., Wilhelm, Synt. Postgate, p. 166, appreciates the magnitude of the difference, but in describing it somewhat exaggerates the rational and intel- lectual aspects of Greek order. * On ‘Spaltung’, ‘Sperrung’, and ‘Hyperbaton’ see Bachhammer, Havers, Spalt., Lindhamer, de Vries, Rass, Kithner, 11, pp. 6ooff., Fraenkel, Jktus, pp. 162ff. 3 Cf. the pertinent criticisms made by Loepfe, p. 10.THE NATURE OF THE PROBLEM 7 few thousand contemporary languages on which we have some information and the handful which we can trace back into the past represent only a tiny portion of the whole field of human language, and that within that portion the variety of structure which we can find is almost as great as the variety which we can imagine, we may well conclude that we have no right to expect anything. Before we make our initial choice from the list of ten categories, let us remind ourselves of some important facts by a glance at the practice of modern spoken English. If we were asked by a foreigner for a general rule of word order in English, we should say, I suppose: The subject precedes the predicate. On second thoughts, we should add: That is, in a statement the subject must precede the predicate; any utterance in which part of the predi- cate, in the form of an auxiliary verb, precedes the subject is a question. We shall have more than second thoughts, but let us consider for a moment the nature of our statement so far as it has gone. Our immediate classification of English sentences is in syntactical terms, according to whether the subject precedes the predicate or the predicate the subject. Yet the determinants of these alternative orders are not themselves syntactical; they are logical. Even then, the rule is not absolute; for there are questions which in respect of order are indistinguishable from statements, e.g., ‘You haven’t done it?’, and the classification of the logical conditions which determine order in questions is subtle and difficult. Furthermore, our statement that the subject precedes the predicate in indicative statements is not entirely true. Consider, for example, ‘Here comes a policeman’, “There goes a bus’, ‘Never have I heard such nonsense’, or ‘He would not do it, nor would I expect him to’. The importance of these exceptions lies in the fact that they are not instances of subordinate syntactical rules concerning negatives or demonstratives or adverbs or conjunctions; they are instances of the occurrence of specified words in specified patterns, and their usage can be communicated to a foreigner only by exhaustive lists and examples. This reminds us that we cannot expect to find that the determi- nants of order in any language are all of one type. Determinants * Cf. Brugmann, Vschd. pp. 1 ff.8 GREEK WORD ORDER of several different types may, as it were, pull a given utterance in different directions, and the order in which it is eventually formu- lated will represent the victory of one type of determinant over the others." The co-existence of determinants of different types is one of the fundamental facts of language. If I construct a sentence containing a subject and a verb which I have never used before—because, let us say, someone has just invented them—my unthinking grasp of a highly general and abstract syntactical rule will ensure that I put the subject before the verb. But this application of a rule to an unfamiliar content does not weaken my adherence to certain familiar patterns which are in conflict with the rule, such as ‘nor would I expect him to’. Historically speaking, over a long enough period the balance of power between determinants demonstrably shifts ;? but it would be a very unusual language in which all the utterances of a given individual speaker were wholly and ex- haustively determined by mutually exclusive rules belonging all to the same type. ‘With these considerations in mind, let us turn back to our Greek example, Mpctoydpos fixel, and make our first choice of a type of determinant for further exploration. It is clear that not all the ten types listed afford an equal purchase. The phonological type offers no firm purchase at all; rhythmically, the word Tpwtoydpos is choriambic; accentually, it is paroxytone with a short penultimate vowel; it begins with a consonant, and ends with one; its vowels are all back-vowels. The number of ways in which Mparteyépas can be seen to differ phonologically from fei is very large, and the possible phonological reasons for the order Tpatorydpas fKe are accordingly so numerous that no obvious starting-point of enquiry suggests itself. Similarly, if the determinant of the order is habit, or ceremonial, or aesthetic variation, it will not normally be discover- able. When the author is dead, there are limits to the inferences we can draw from what he has left us in writing. The fact that some types of answer are hard to obtain does not mean that these answers are wrong; it means only that we shall be * Schwyzer, Gr. 11, p. 690. Firbas, Comm. pp. 39f., WO, p. 73 speaks of ‘hierarchies of principles’ and illustrates the growth of the domination of syntactical principles of order in English. ? Bloomfield, pp. 156f.THE NATURE OF THE PROBLEM 9 unwise to attempt such an answer until we have first explored the possibilities of easier types. The possibility of an answer in lexical or semantic terms depends in the first instance on observation and thereafter on statistical compilation or negative observation. The process reveals that there are many words in Greek which never begin a sentence, and others which never end one. In addition, there are some which are disproportionately common at the beginning of a sentence, These data are useful, but they do not help us with Tlpoxtacyépers FiKet. An answer in morphological or syntactical terms stands at the opposite extreme to the phonological. Morphological and syntac- tical categories in Greek are few, and the possible alternative rules of order in terms of these categories are correspondingly few. Also, there is seldom room for disagreement on whether a given word is noun or verb, subject or object or predicate.’ The a priori advantage of a morphological or syntactical answer is strengthened by the fact that in very many languages the determinants of order are syntac- tical? and may be thought to be strengthened by ancient theory. The order subject—verb is described by Demetrius Eloc. 199 as ‘natural’ (A puoiKi t&Eis), and noun—-verb by Dionysius Comp. Verb. 5 as Ti pUoel érdpevov. But any inclination which the statistician may feel to welcome the ancient critics as allies may falter when he considers their reasons. Demetrius, speaking specifically of narra- tive, says that the subject-matter (16 tepl o¥) of a sentence should be stated first, and 6 totré tomw second, which is not quite the same as saying that the syntactical subject precedes the syntactical predi- cate. He continues: ‘but, of course, the order can be reversed... . Ido not unconditionally approve the former order and condemn the latter; I am merely setting forth 1d quoikdv elBos tis TéGecs.” Demetrius’s conception of ‘nature’ is not necessarily based on linguistic usage, as we may see from Dionysius. Dionysius thinks it natural that nouns should precede verbs because nouns denote substance (ovc{x) whereas verbs denote accident (ouuPefnkds) and substance is ‘by nature’ prior to accident. He once believed, he says, that in order to write well we should follow nature and put nouns before verbs; but he realised later that this a priori philoso- * The problem of Greek order in general has usually been discussed in syntactical terms, e.g. by Chantraine, Frisk, Kithner and Schwyzer. ? Barth, pp. 22ff.Io GREEK WORD ORDER phical view of the relation between substance and accident was irrelevant to good writing. He deserves credit for his observation, _.- and gratitude for the honesty with which he reveals the divergence of interest between metaphysics and linguistics. When we come to examine syntactical statistics, we shall do so without philosophical allies. (iii) MATERIAL It is proposed in the next three chapters to identify types of determi- nant in Greek word order; it is therefore necessary to define ‘Greek’ for this limited purpose. Clearly our enquiry must be into early Greek; the study of word order in the Gospels or Plutarch is interesting and worth-while, but no one will want to turn first to them for an answer to a question about the structure of the Greek language so long as the material of earlier centuries is available.t Equally clearly, our attention must be concentrated on prose rather than poetry. In poetry rhythm is among the determinants of form;* and where the determinants of form generally are more numerous, the isolation of the deter- minants of word order in particular is likely to be more difficult. This consideration is responsible also for my comparative neglect of that large portion of fourth-century prose in which rhyth- mical and phonetic preoccupations are demonstrably at work among the forces determining the form in which authors express themselves. I have therefore paid special attention to Herodotus and docu- mentary inscriptions, I am aware that Herodotus was an artist, the power and beauty of whose work are manifest onevery page; yet there are two important respects in which his language is ‘natural’. He does not try, as Thucydides constantly tries, to say too much in too few words; nor does he wrestle clumsily with language in the manner of the so-called ‘Old Oligarch’. Secondly, the rhythmical and * The statistics in Frisk, pp. 16ff., show important differences in syntactical order between Hellenistic and Classical Greek; we may perhaps come to the conclusion that the major difference is the emergence in Hellenistic Greek of syntactical determinants which are irrelevant to Classical Greek. ? Porzig, p. 97, says that the exigencies of metre do not create new patterns of order but only determine the author’s choice between existing alternatives. Tam not sure that this distinction does not beg the question; cf. Wackernagel, Dich. pp. 6f.; Bloomfield, pp. 157, 165; Spiegel, p. 514.THE NATURE OF THE PROBLEM Il phonetic considerations conspicuous in the fourth century are not in evidence in Herodotus.?_, As for inscriptions, it must be admitted that Greek documentary inscriptions exhibit, at certain times and places, positive charac- teristics of their own, and that we cannot assume without qualifica- tion that their language is ‘basic’ or ‘natural’ Greek. Nevertheless, these characteristics are few and easily identifiable; the establish- ment of formulae, even at Athens, was remarkably late; and there are several other considerations which make inscriptions of prime importance to the student of word order. First, the manuscripts and papyri of a prose literary text reveal just enough differences in respect of order to make us reluctant to rely too heavily on the authenticity of the order in any one passage, whereas an inscription may be presumed to be only one stage removed from an autograph, and slips of the chisel are likely to be rarer than slips of the pen.? Secondly, the composition of documentary inscriptions is moti- vated by a desire to communicate clearly and not to invite admira- tion as a work of art.3 I have therefore chosen several of my documentary examples from places which were not centres of cul- tural and literary development. I cannot swear that the decrees of the Ozolian Locrians do not betray the hand of a mute inglorious Gorgias, but I may be allowed to doubt that and to believe that in early documents from the Peloponnese, the North-West and Crete the influence of Kunstprosa is minimal. There is some value in applying to word order, and to all problems in the history of the Greek language, a principle which was applied on a famous occasion to a matter of higher importance: mpd@tov tv Tats TéAEon grTHCCpEV Trotév ti éomiv. 1 Herodotus’s language undeniably shows some characteristics which we associate with spoken rather than with written composition (cf. Pohlenz, Herodot (Leipzig, 1937), pp. 209f.), and this makes analysis of the logical relations between the elements of a Herodotean utterance easier and more profitable (cf. Loepfe, p. 59). * Striking and characteristic patterns of order in literature are always in danger of over-systematic emendation, from which inscriptions are com- paratively safe (Schéne, Umser. pp. 145 ff.). 3 Schick, pp. 362f.CHAPTER II LEXICAL AND SEMANTIC DETERMINANTS (i) POSTPOSITIVES AND PREPOSITIVES Most Greek words are ‘mobile’ (symbol ), in the sense that any one of them may be found at the beginning of a clause, at its end, or in the middle. There are, however, two important categories of limited mobility. These are: (1) Postpositives (symbol g). These words never, or only in certain specifiable circumstances, occur at the beginning of a clause. Some of them are unconditionally g, ie. may not in any circum- stances begin a clause. They are: (@ The particles &per ($a), a8, yp, ys, Bal, 8, Sine, Onv, uy, pty (uav), ov (dv), ep, Te. Gi) The word to1, which may be regarded either as a pronoun or as a particle, according to dialect. Gii) The pronouns pe, Hou (ev), Hol, LIV (viv), oge. (iv) The modal particle év and its synonyms ke and Ka. Certain others may conveniently be treated as unconditionally ¢, since the circumstances in which they may begin a clause are extremely restricted and the instances very few: (v) 8h. (vi) The pronouns of and og. (vii) The oblique cases of atrés in the anaphoric sense. In the sense ‘self’ they are /; but although this semantic difference might be expected to cause trouble, I know of no instance of initial atrr-dv, etc. which requires to be translated as anaphoric and forbids the translation ‘self’. The status of other q is less sharply defined: (viii) The rule that the indefinite pronouns, adjectives and adverbs Tis, 1109s, etc. are g is true on the whole, but to some extent artificially protected by conventions of accent and translation. Presented with tis évS0v; as a complete utterance, we should write an acute accent on tis and translate ‘who is in?’ even when ‘is someone in?’ might seem more appropriate to the context." *Ts A. 4g. 1344 siya ths TANyt dirtel Konpleos otrracpévos; really a question?LEXICAL AND SEMANTIC DETERMINANTS 13 (ix) The pronouns oe, cou, coi, when initial, are by convention accented and are translated-as emphatic. When medial or final, they are written as enclitic if the context permits their translation as unemphatic but accented if it requires them to be translated as emphatic. The statement ‘unemphatic oe, cov, coi are enclitic and postpositive’ thus incurs the danger of circularity inherent in the concept ‘emphasis’. (x) Similar difficulties beset the classification of viv/viv, the oblique cases of ogeis, the present tense of pdvon, and the oblique cases of tueis and Wueis, though the two last lack the complication of accentual conventions. . (xi) The verb elven cannot be classed as ¢ without many qualifica- tions and reservations, not all of which admit of a satisfactory classification. elven as copula tends, in most authors, to be treated as q; when it is first word in a clause, we import into its translation nuances which the context does not always demand and sometimes scarcely justifies, I use the symbol MM for elven in its copulative sense. In generalising about g it is the unconditional g, categories (iH vii), which I have in mind, and it is the use of these words especially which my examples are designed to illustrate. The definition of g as words which never begin a clause necessi- tates a modification of our definition of ‘clause’. One such modifica tion is discussed below. The other is that vocatives, oaths and parentheses, although often marked off by commas in modern texts, are often immediately followed by ¢, e.g. Ar. Pax 137 GAN’ & pen” &v por orttaov SrmAdv eer; Pl. Prm. 1378 tis ov, elteiv, yor écrro- xpiveiten; This amounts to saying that in respect of g vocatives, etc. do not affect the order.* (II) Prepositives (symbol p).? These words never, or only in certain specifiable circumstances, end a clause. They are: G) dAdG, crdp (air&p), f, 7, Kel, OBE (UNBE), otrre (et) elre. ii) u\=‘lest’. (iii) Relatives, including the indeclinables et, érrel, Iva, ¢ etc, (iv) 6=‘the’. In other senses 4 is not p. * Wackernagel, Gesetz, pp. 391f.; Kieckers, V7. Sag. p. 150. The most remarkable postponement of gis ina v./. at Ar. Mu. 398f. Kol Tréds, & wdpe ov Kai Kpovicov Sgcov Kal PexkectAnve, eftrep B&AAst Tous EridpKous, Sf[T” ovxXl Lipeov’ Everproev; (f\t" V: és codd, plerique: més Sit’ contra metrum Ral.). See also p. 15 n. 1, p. 16 n. 3. ? Ammann, Unt. 1, p. 12.14 GREEK WORD ORDER (v) Prepositions. In this category considerable historical changes and differences between genres are observable, but in prose generally .. all the prepositions are p except sometimes rept; &veu, Evexc, xcpiv, opis, etc., reasonably classified as ‘improper’ prepositions, do not come under consideration here. The simple negative is not easy to classify. It has obvious p characteristics; it might be argued that expressions such as ef 5 pr} or ade piv dv yévorto, exeiveos 8° of are just as elliptic as e.g. otx of8’ 6trou, and therefore do not tell against its inclusion among p. Since, however, it may constitute a complete utterance by itself, and since the types of clause in which it may appear last are numerous, I do not treat it as p. The three types of minimal complete utterance are |||, ||p/|| and ||Mq||. By definition, ||p |, |I¢||, ||44p || and |7/4 || are impos- sible. When two M and one g constitute a complete utterance, the two arrangements which are theoretically possible are || Mg M\| and ||44 .Mq||; with two M and two q, there are three possible orders, \\Mog AA, Mg 2Mg\\ and |Mf Mgq||s with three M and two g, six possible orders; and so on. Yet these alternatives are not equally used. Greek has a strongly marked preference for || Mg(q...) M (M...), accumulating ¢ after the first 14.‘ This is an observable historical fact which could never have been deduced from the definitions of M, p and g. In this pattern word-groups which we, as speakers of modern European languages, would regard as indis- soluble are disrupted, e.g. Callinus, fr. 1, 20 doTrep yap wiv TrUpyov tv dqGoApoioww dpdaiv, Ar. Nu. 257 dotrep pe Tov "APdpave’ Sires bh Quote, DGE, 412, 3 (Olympia, VI 3.c.) & 5€ Ka fpétpa & Sapoola teAla elé SixdSdo0, Hdt. vi, 63, 2 &v S€ ol Adooovi xodvep. . . 4 yor att tikra. Similarly, words which, as we should say, ‘go together in sense’ are widely separated,? e.g. Lys. 1, 17 Tatité& pou mavra els thy yveouny eloner; in particular, g which ‘belong’ in patticipial and infinitive clauses are commonly associated with the words to which those clauses are subordinate, e.g. Hom. JI. xx1, 347 xadper 5 piv Goris eeipy (a rare type), Hdt. 1, 30, 4 Kal op: elBe Grraat Téxva éxyevoueva, D. |, 18 exéAevod por attov vourras uiobdd- gacfai. Hence a sentence such as Pl. Euthyphro 6a pioei tls pe * Wackernagel, Gesetz; Schwyzer, Gr. 1, p. 692; Leumann, pp. 611 ff.; Delbritck, Altind. pp. 47£., 59. * Cf. Brugmann, Vgi. pp. 681ff.; Delbriick, Vgi. pp. 4off.LEXICAL AND SEMANTIC DETERMINANTS 1g #€apaprévelv is to be treated as only an apparent example, not a real example, of a q (ue) beginning a clause." Generally speaking, the earlier the Greek, the more closely does M+M+q>MqM approximate to a rule. It is not, however, a question of certain metrically convenient patterns establishing themselves in the dactylic hexameter and influencing the subsequent development of the language. The phenomenon occurs in other Indo-European languages, especially in Indo-Iranian, eg. Old Persian, XPa (Xerxes, Persepolis) 15 uta-maiy tya pita akunaus = kat pou 6m1 (6) trathp éroincey.? The progressive tendency in post-Homeric Greek to distribute ¢ within the clause, instead of concentrating them after the leading M, is a secondary phenomenon. One of its principal causes is a certain development, independent of anything we have discussed so far, in the treatment of p. Even in the earliest Greek there exists a tendency to arrange an utterance consisting of M, p and q in the order pMq, and this tendency is increasingly strengthened in historical times. This, again, is an observable fact which could not have been deduced from definitions. Its extreme development may be seen in the evolution of ‘compound verbs’. By the end of the fifth century prefix and verb were virtually indissoluble in the language of prose, and were presumably already an accentual unit. This evolution may be symbolised M,9+M,>M,g. Contemporaneously with the coalescence of p and M, certain words underwent semantic changes and declined from M-status to p-status. The demonstrative pronoun 6 in Homeric Greek may be symbolised MJ’, i.e. M destined to become p (and, of course, already in process of change even in Homer). The coalescence of p and M is most clearly seen in cases of pMq where g is a connecting particle, e.g. Hom. J. xxiv, 665 TH Sexérq BE Ke Odrrroipev, Hdt. rx, 49, 3 cd ToU ToTapod yap ogi otk zEfiv Sep poptectn, Ar. Nu. 1198 Strep of trpotévOan yap Soxoticl pot trofelv. These three examples may be symbolised: Hom., pMqqg M; * The MS. text of E. [4 1435 twatioal pe pt Kéxize, accepted by Wacker- nagel, Geserz, p. 392, might be tolerable if the context permitted (as it does not) the translation ‘stop abusing me’. Occasionally the inversion of Mf and q normal in sentences of the type pnot ue tfapaptdvely is found where some degree of pause immediately before q seems inescapable, e.g. Lys. x11, 10 Gpoorv. . .AaPdv 1d TaAavtév ps doe. Cf. p.13n.1 and p. 16n. 3. 3 Kent, pp. 96f.; Wackernagel, Gesetz, p. 404.16 GREEK WORD ORDER Hdt., ppMgqg MM M M; At., ppMg Mg M. More commonly, the pM complex may be disrupted by connecting particles, but not__. by other g, e.g. Pl. Prt. 310¢ 6 yép Tot Tats ye & Lérrupos érréSpar, where the pattern is pggMq pM M. This pattern is typical. The placing of particles after the definite article or preposition is so fundamental a characteristic of Greek, and we learn it so early in our study of the language, that we do not reflect often enough how curious a phenomenon it is. It is in fact the result of a compromise between pattern and principle. On the one hand, there is the increasing tendency to treat pM as an indissoluble unit; on the other hand, there is a poetic language in which pg(g...) M is unrestricted and the evolution M*>p only in its initial stages. Symbolically, pg, Mg, where g, is a connecting particle and g, any other g, is the joint product of pg(q...) M, M*q(q...) M, and PMqq...)." As we should expect, the treatment of p+M+g varies greatly from one dialect to another and from one period to another. Herodotus, for example, inserts a g pronoun between article and noun or between preposition and noun; in Attic prose this is very rare with yo, even rarer with gol, and unexampled with those cases of avtév which are equivalent to uty, of, ogi. Kat por and GAA& por with imperatives survive more strongly in Attic than any other kal gM or ddAK gM where gis a pronoun. A literal translation of DGE, 412, 3 into Attic—To 8’ dv yrigicua Td Snudotov KUpiov Ein Sixdzov— although not entirely unparalleled,? would be extremely unusual. Thucydides commonly places anaphoric ovt- immediately after connective xal; this is not favoured by fourth-century prose. On the other hand, fourth-century writers arrange the words in certain parentheses to yield M(q)(|)gM!... in preference to M(q)|_Mq..., eg. D.1, 19 ‘ti otv;’ &v Tis ettrot, ‘oW ypdgeis ToT’ elvoa oTparrice~ miké;’, Pl. Euthd. 297¢ veworl, yor Soxeiv, KerromrenAeuKéri.3 The deferment of connecting particles also provides evidence of the formation of other types of complex unit. One common type con- sists of a word repeated in polyptosis, or a pair of closely related * Cf. Leo, p. 419. ? Eg. Th. vi, 64, 1 Tos yap dv ious. . tos Irrméas. . . Adare &v usyéAa. For other types of pgM which are unevenly distributed among dialects cf. Wackernagel, Anr, pp. 27 ff. (on tav Tis and al tis Ka) and Wilhelm, Beitr. pp. 44f. (on t&v Tg ToArtOv). 3 Wackernagel, Gesetz, pp. 392, 397. Cf. p. 13m. 1, pe 15 nea.LEXICAL AND SEMANTIC DETERMINANTS 17 words, e.g, A. Eu. 529 GAA’ AAG B’ Epopever. This accords with such phenomena as Tope piAgs gfe and Urrtp atrrés arto." A numeral and the word to which it refers are sometimes treated as indissoluble,” e.g. SEG, x, 2 (Marathon, VI/V B.c.), 19 ff. Tpidxovta avEp[as 58] Tov &ySva Emdpoaobal, Complex units account for many clauses in which g are not placed immediately after the leading M, but there is another phenomenon of much wider significance. Contrast Hdt. 1, 10, 2 kal f) yuvth trrop& uly efidvrx with 12, 1 Kod piv Exelvn eyxeip{Siov Soticn Korra- kpUrrret. In 12, 1 the pattern pg... is normal; in 10, 2 we have ppM MqM. Now the idea that Kal *| yuvt érop& constitute a complex unit cannot be taken seriously; and the fact that in reading the narrative slowly and clearly we would make a slight pause between yuvi and érop% is highly relevant. The example suggests that g are not necessarily placed after the leading p or M of what would traditionally be defined as a ‘clause’, but may occupy a similar position within one of the word-groups which constitute the clause; a ‘word-group’ being recognisable by the pauses of the voice which precede and follow it.3 This suggestion is supported by very many passages, e.g. Th. vi, 9, 3 Kal pds pév Tous tpdrrous Tous Uyetépous dobeviis dv ou & Adyos ein. The words kal... Upetépous are not a ‘clause’, but they are a well-defined word- group; doGevts begins the next group, which takes the familiar form Mg pM Mt. A word-group, defined in terms of natural pause, may be very short,# as in Hdt. 1, 2, 1 tora pév 84 toa pds Iga ogi yevéobau (note that ia pds oa are treated, as we should expect, as a complex unit), Lys. 11, 2 Spoos 88 6 pev Adyos pot trep! Torrey, 6 8’ &ydov KTA. Indeed, circumstances often arise in which we may wonder whether the first word of a clause is to be followed by a slight pause or is simply equivalent to p, e.g. Pl. Pre. 3338 otxotv év &v ein t ocoppoawvn Kal t opla. One effect of breaking up a clause into word-groups is to distri- bute g; and one consequence of this distribution is that g are often to * Cf. Haupt, pp. 184ff.; Schulze mentions the phenomenon in Lithuanian. ? Cf. Fraenkel, Zktus, pp. 120ff. 3 Fraenkel, Kolon, pp. 319ff.; de Vries, pp. 20f. See also p. 19 n. 1. 4 “Kurzkola’ in Fraenkel, Kolon, pp. 327, 343, al. Section (iii) below is relevant. 2 DG18 GREEK WORD ORDER be found in close proximity to the words with which, as we should say, they ‘go’.' As speakers of modern English, we find PI. _.- Euthyphro 9c totrrov pty dglnut oe and 15 £ viv yép otrevBw Trot more ‘natural’ than toUtov pév 0” donut and viv yap trot oTrevBe. We are therefore bound to consider the possibility that the funda- mental reason for the distribution of g and the consequent departure from the original tendency to collect g after the leading M of the clause is not the fact that a clause can often be broken up into word- groups but a desire to bring closer together words which ‘go together’ in sense. We might suspect that we see the result of a conflict between pattern and sense in those clauses in which q is repeated, e.g. E. Md. 250f. cs tpls av Trap’ atria oTfivan GéAorp” &yv paAAov # texeiv raf (pMg pM M Mq M pM M). & is the q most commonly repeated,’ but others may be,3 e.g. X. Cyr. 1v, 5, 29 oxéyar St of évmi pot Tepl of olos dv trepl epé Errert& por uépoH (Mq|||M Mg pM|\|.M M+ pM||\Mq M). This phenomenon is not confined to Greek, but occurs also in Vedic Sanskrit, e.g. Reveda 1, 35, 11 tebhir no adya pathibhih sugebhi raksa ca no adhi ca brithi =Tonrrais hes THuepov S8ois eUPdrrois os Zé Te Hyds Urvép Te (Hudv) é&yépeve, where the pronoun no ( = I Pl. 18 5 o- = 4 (xiii) réte Hat. 3 I o- = 2 Lys. 3 1 o—- = I (xiv) ontrrés Hat. 15 6 =—- = 4 Lys. 15 2 = 8 5 . PL 21 30 4 26 (xv) 6 ontrtés Hdt. 3 to- = I Lys. 4 20- 1 3 Pl. 4 I 2 3 (xvi) GAAos Hat. 6 4 3 °0- 3 Lys. 8 2 = I 2 Pi. 32.3 7 2LEXICAL AND SEMANTIC DETERMINANTS 23 | I rt mi vv (xvii) Etepos Hd og Lys. 4 2 2 I 5 Pi. _ I — I (xviii) &ppétepor Hdt. 2 — -—- = 2 Lys. 4 - - - = Pl. - - ro- = (xix) TTOAUs, etc. Hdt. 6 —- -—- = 4 Lys. 26 7 2 1 8 Pl. 7 4 3 °- WW (xx) TOAAGKis Hat. I - = - —- Lys. 20- =| = = Pl. 305 0°5>O0~€~«S I (xxi) els Hdt. 3 05> > ~°25 3 , Lys. I 30- - Pi. 3 300 2 I Tt will be seen that there are differences between authors; the otros group is overwhelmingly J/* in Lysias, much less so in Herodotus and Plato. Greater differences are apparent in: (xxii) 8c, T75e, GBe, toidcbe, toodabe. (xadii) és and adverbs formed with the stem twovt-. I ff W@W iVvVv (xxii) 68e, etc. Hdt. 3 °- I 2 Lys. 3 0- - = I Pl. Ir 2 2 2 23 (xxiii) Tré&s Hat. 7 20- 2 2 ly. 13 #4 —-— — 4 Pl, 20 4 4 — 10 Trev TOS Hdt. 2 0— —- = = Lys. Io- = >= = Pl. 4 - Io I Of the instances of 55, etc. in column V, 16 out of 17 in Herodotus and 11 out of 23 in Plato actually occur as the Jest word of a clause.24 GREEK WORD ORDER Semantically, 4* do not constitute a well-defined class. One group has a purely demonstrative character,’ e.g. éxelvos, and a_ second group consists of logical connectives,? e.g. Syas. Others have affinities with one or other of those two groups, as éy, au, etc. have with the demonstratives or Tévtes with the logical connec- tives. Others have affinities with both, e.g. obtws and viv;3 and others again with neither, e.g. ToAUs.4 * Kieckers, Se. Vds. pp. 18ff., r10ff., 126f.; Or. R.1, p. 93 Kaibel, pp. 99f.; Bergaigne, p. 176. For statistics of pronominal M* in relative sé., in terms of Sand P, see Frisk, p. 39. * Bloch, passim. 3 Ammann, Unt. 1, p. 42. 4 Ammann, Unt. 1, p. 16. I have not exhausted the list of M*; see Spiegel, p. 514, Kieckers, Vb. Sag. pp. 145 ff., Or. R. 1, p. 9, Schéne, Vschr. on ‘say’, and Bloch, pp. 243 ff., Kieckers, St. Vas. pp. s0ff., on ‘be’. See also ch. tv, 8 (iv) below..- CHAPTER III SYNTACTICAL DETERMINANTS (i) GENERAL PRINCIPLES In Greek prose of the fifth and fourth centuries B.c. the subject (S) tends to precede its verb (V); the relevant statistics can be broken down to show, for example, that in main clauses in Xenophon, HG the ratio SV’: VS is 4:0, in relative clauses in Lysias 9:0, in temporal clauses in Herodotus 1-2, and so on. The ratio OV: VO shows greater fluctuation, the extremes being 0-7 and 4:0, but in most types of clause in most authors it exceeds 1:0." These ratios seem to justify us prima facie in giving SV and OV the status of syntactical rules, The problem is then to account for the instances of VS and VO, and this problem we might expect to solve by finding conditions which are present in all instances of VS or VO but absent from those of SV and OV, or, if present in SV and OV, are counteracted by other conditions absent from VS and VO... and so on.” If we confined ourselves to purely linguistic conditions, we might imagine that the differences of order between Lys. x1x, 50 Twovete...cos Arotivos Exo1 TéAavTa TeTtapéKovTa TAEio # Soc adds pordyer and 51 é&ravteov éxnkodtwv ST TeTtapdKovta téAavta Exo Aidtiwos or between Pl. Jon 5383 ti Bt St étav “Ounpos Afyn ads. . “ExoprSn. . .8{Sco1 and 538c tI 58 Stav Ayn “Ounpos ‘1 S& poAuBSalvy IkéAn KTA.’ are the product of the differences between the ‘constructions’ of the contexts of the words concerned, or that between Hdt. 11, 67, 1 ob yap fv of doqoAts KayBuoew teteAcuTnKdtos paver tov Kipou uldv drodcAekévon ovroxeip{n and 67, 2 & BE St udcyos TeAeuTI}OOVTOS Kay Bloew &be6d5 eBaciAeuce a difference between the perfect and aorist aspects. These hypotheses do not in fact survive testing, and we are constantly confronted with differences for the explanation of which we would need to draw linguistic distinctions of increasing complexity and increasingly obvious irrelevance, e.g. X. HG, 1, 2, 16 AAKiPi&Ens BE * Relevant statistics (in terms of S, O and P=Predicate) are to be found in Frisk, pp. 16ff. ? Delbriick, Germ. p. 10, calls the order SV ‘normal’ and VS ‘invertiert’, Frisk, pp. 39, ad, ‘gerade Stellung’ and ‘Inversion’.26 GREEK WORD ORDER 8Slooxev Excov Tous Te tras Kal Td&v STrArTéSv elkoor Kad éxerrdv, ddv Fipxe MévavBpos, An. 1, 7, 11 GAAOL 8 Foav Efaxiox {Aion frrreis, Ov |. *Aptayétpons Aipxe, Th. 11, 30, 1 Kal’ Aotoxdy, fis Etiapyos trupévvel, AaPdvtes. . .TpOcETToI\aavTo, 80, 6 Opéorten 5é x{Ato1, dv éBaotAcuev *Avtloxos, yeTé Tapavatcov fuveotpatevovto. In short, the relation- ships S-V and O-V are in the last resort no more stable in order than the relationships illustrated in ch. 1 (i). Yet in recognising this fact we must also account for the statistics. What exactly do they prove? Using throughout the classification of clauses explained in ch. 1 (ii), T omit from all the statistics in this section all S—V relationships in which V is elvoa, whether copulative or existential, or copulative ylyveoGou.' I include in O the genitive or dative with verbs which never or rarely take an external accusative, e.g. XpfjoGo1, ém@unpelv, and the dative of the indirect object with verbs meaning ‘give’, ‘say’, etc., where no direct object is expressed. Where any of these verbs do have an accusative object, I treat that as O. I exclude from O the neuter accusative singular or plural of an adjective without the article, e.g. kaxc ppoveiv, Sevdv Troieioban, (ii) PREFERENTIAL WORDS Tt is obvious that when S is an interrogative the ratio SV: VS will be very high; so will OV’: VO when O is an interrogative.? We should therefore expect in general that when S or O is M+, SV and OV will be commoner than when S or O is M’. This expectation may be tested statistically; I give below the figures for S=M* and O=M*. By ‘S=M*’ I mean that S is either (i) wholly composed of M’, e.g. yc 46, or (ii) contains M? and is wholly on one side of the verb or the other, e.g. Tévde tov GvBpa dpé; cases such as tov dévBpa dpdd tév6e are excluded from these statistics. ‘O=/M*’ is to be similarly interpreted. I do not regard ovrrés by itself as ever constituting S or O, and I admit ws by itself as constituting S or O only when it requires the translation ‘everyone’ or ‘everything’ and forbids the translation ‘all of them’, ‘all of us’, etc. 7 Ebeling’s statistics of the copula give a picture quite different from that which is given by statistics of verbs other than the copula. ? Thomson analyses the logical circumstances in which interrogatives are displaced from their normal leading position.SYNTACTICAL DETERMINANTS SV tya, etc. 6 cnrés GAAS Gupdtepor TrOAUS els Torats Ratios Ht. Lys. Pi. Hdt. Lys. Pl. Hat. Lys. Pl. Hdt. Lys. Pi. Hdt. Lys. Pl. Hdt. Lys. Pl. Hdt. Lys. Pl. Hdt. Lys. Lys. Hat. Lys. Pl. Hdt. Pl. Hdt. Lys. Pl. Total Hdt. Lys. Pl af 9 49 5 12 40 21 25 21 2 HM NON 0 | wae enw ennn | 44 118 225 Pldatl il tle lb bil bla nn nw wha Slliwlng BBw 5 SVIVS 44 315 3°93 3°58 OF VO 5 2 2 2 15 3 28 25 46 7 7° 19 I — 8 I 2 3 2 _ 3 I 6 _— 4 — I _ 2 — 2 I 12 4 I _ 4 — I _ I _ 19 — 7 I I — I — 45 27 88 Ss Ir 31 244 6 OV:VO 1°67 828 GREEK WORD ORDER As these figures accord closely with the general picture of M+ given in I (iii), so too the figures for 65¢ and was reflect the _. differences between authors illustrated there: SV VS OF FO o5e Hdt. _ I I qI Lys. I _ _— I Pl. 7 2 8 2 Tres Hat. — 3 7 5 Lys. - - 11 I Pl. 7 - 8 2 (iii) ORDINARY WORDS Let us now see what happens when S and O are M°. I distinguish between: (a) | SV, in which no M of any kind precedes S$ in the clause. (6) nSV, in which S is preceded by a negative but by no other M. (c) -SV, in which S is preceded by at least one M other than a negative. So too nVS, -VS, | OV, etc. Clauses containing S, O and V will appear twice in the tables, once for their S—V’ and again for their O-F relationship. Thus e.g. Pl. La. 184 vow & Thy évavtiav. .. Adéxns Nixie {eto is both -SV and -OV. I give separate figures for seven of the eight types of clause defined in II (ii); a/inf- is so rare in the texts analysed that I have omitted it. mn. sb. pt.* pt. alpt. inf.* inf. |Sh Hadt. 27° «15 «15 G4 3 > - WW Lys. 15 12 5 6 —-— — WW Pl. 9 15 3 40> 6 11 nSV Hdt. - -— T roS— I Lys. I ro-— — se Se Pl. I I Iro- > = I SV Hdt. 5 305° ~°*2 Lys. 3 3 Io>- > 2 Pl. 3 2 2 2 = I[vs aVS [OV nOV |vO nVO -VO SYNTACTICAL DETERMINANTS Hdt. Lys. Pi. Hdt. Lys. Pi. Hdt. Lys. Pl. Hdt. Lys. Pi. Hdt. Lys. Pl. Hat. Lys. Pi. Hadt. Lys. Pl. Hat. Lys. Pl. Hdt. Lys, PL. Cans 12 22 14 34 Il 17 2 15 4 10 il | oba wlan OvwOD ANY WH = HeN » N= The totals for the three texts are: Hat. Lys. Pi. SV 99 60 63 VS 68 12 22 pt. * 12 pte 13 26 yw 10 35 17 Il vow vNw OV 76 178 89 alpt. inf* a 2 = _ I ro ro I I 30 30 2 -— — I roo ro 2 - vO 127 57 79 14 39 22 4 NVN YAK | wow NAD30 GREEK WORD ORDER The differences of ratio between the four principal types of clause (omitting pt.*, a/pt., inf.* and a/inf. as inadequately represented) are: _ Hadt. Lys. Pl. SV:VS OV:VO SV:VS OV:VO SV:VS OV:VO mn. rig 059 4°75 4 2:6 1:07 sb, 1-97 0°53 16 6-75 18 158 pt 125 037 15 195 3 067 inf. 3°25 1:27 <0 281 65 1°32 It appears that there is a consistent preference for SV, and that this preference is more marked in inf. than elsewhere. With that exception, the most conspicuous feature of the tables is the differences between authors.’ These differences, however, are not consistent differences between dialects, periods and genres; authors are not even always consistent with themselves. In Hdt. vim, 1-48, Lys. x1x (De Aristophanis Bonis) and PI. Jon, the comparable totals are: SV VS Or vO Hdt. 7 45 85 76 Lys. 38 24 61 55 Pl. 42 21 43 33 Comparing now the ratios SV: VS and OV: VO for all six texts, we have: ve SV:VS— OV:¥O Hdt. mt 146 o6 Hdt. viii 1°86 12 Lys. x 50 312 Lys. x1x 1°58 Weir Pl. Laches 2:86 113 Pl. Zon 20 1130 * Further data illustrating authors’ preferences in different types of clause may be found in Behaghel, Sz. V3. p. 280, and in Frisk, pp. 28ff. Hermann, pp. soof., points out that there are no structural characteristics of sb. which are general JE. ‘Nachsitze’ (see p. 20 n. 1) which fulfil the requirements of my argument by containing either S=M? and V or O=M? and V are too few in the texts analysed to be distinguished from other mn. So far as they go, they exhibit S V, VS, O V and VO.SYNTACTICAL DETERMINANTS 31 Plato is the most consistent. Lysias shows a very much greater preference for SV and OV in x1 than in x1x. Herodotus in vit agrees with Lysias and Plato in preferring OV, but differs greatly from his own practice in 11. It is clear that these statistics are very far indeed from estab- lishing for ‘Classical Greek’ simpliciter anything worth calling a syntactical rule of word order. Extended to a much greater variety of authors and texts, they would no doubt give us an interesting picture of the vagaries of individual preference—and thereby suggest with increasing force that all patterns of order which are describable in syntactical terms are secondary phenomena.CHAPTER IV LOGICAL DETERMINANTS A. GENERAL PRINCIPLES (i) EMPHASIS The fact that Greek utterances identical not merely in structure but also in content may still differ in order, the variations in structural preference between authors and between different portions of the same author’s work, and the high proportion of ‘abnormality’ even in authors whose preferences are consistent, suggest that some, at least, of the determinants of order must be sought not inside the utterance itself, but outside it, in its relation to its context.t In modern spoken English such relations are expressed by modification of the tone and volume of the voice, so that two utterances which in writing are identical may be revealed in speech as standing in quite different logical relations to their contexts.” These relations in Greek have sometimes been discussed in terms of ‘emphasis’, sometimes in terms of the distinction between ‘logical subject’ and ‘logical predicate’. For pedagogic purposes, rough rules have been formulated in the terminology of emphasis, e.g, ‘the emphatic positions in a Greek sentence are at the beginning and the end’.3 Yet the term ‘emphasis’ is for a variety of reasons unsatisfactory. (a) ‘Emphatic’ is commonly used to describe both words which are the focus of the speaker’s emotion and words which are essential to the clarity of his argument.* Some passages of * Cf. Loepfe, pp. 8, 130, Kaibel, p. 96, on the failure of purely grammatical explanations. 2 Cf. Brugmann, Vschd. pp. 4f., 9ff., Richter, p. 28, on the variety of function fulfilled by tone and volume of voice. 3 Denniston, Pr. St. p. 44, Delbriick, Vgl. Synt. pp. 110ff. 4 Firbas, Comm. p. 39, uses ‘emotive’ to denote all kinds of ‘emphasis’. Yet the distinction is vital; clear explanation and the stimulation of emotion are often incompatible; so are humour and explanation, for the humour of the ‘unexpected requires the speaker to create a misleading expectation, and this is not a common or profitable didactic technique. Richter, p. 37, makes a fundamental distinction between ‘geftihlerregende Rede’ and ‘berichtende Rede’.LOGICAL DETERMINANTS 33 Greek prose are designed to stimulate in the hearer pity, terror, anger, scorn or pride;.the majority are not: they are designed to make the hearer understand, and the only emotion which sustains them is the determination to communicate intelligibly. The difference in purpose and circumstances between language which stimulates irrational emotion and the language of ex- position is profound, but the term ‘emphasis’ obscures this difference. (6) Emphasis is necessarily a matter of degree. There are some short utterances in which it is possible to designate one component unhesitatingly as ‘emphatic’ and another as ‘unemphatic’; but most Greek utterances are longer and more complex. Consider for example Heraclitus, fr. 57: S18éoKoAos 5 TAcloTav ‘Hoiobos: tottov yap eriotavtai TAsiote elSévon, Sotis Hugonv Kal eUppdvny otk tyivoooxev: tov yap Ev. If we possessed the original context of this fragment, we should probably be able to say which of the words in the first clause is less emphatic than its neighbours; but with that exception, I should find it very hard either to allocate the remaining words between the categories ‘emphatic’ and ‘unemphatic’ or to arrange them on any scale of emphasis." Objections of a similar kind may be brought against all analyses of logical relationships; but we must seek, and may find, an analysis which admits of more absolutes than the concept ‘emphasis’ does, (c) To a remarkable extent, individuals may disagree about the location of ‘emphasis’ in a given passage of Greek,? and an indivi- dual may disagree with himself on different occasions. (d) These three defects of analysis in terms of ‘emphasis’ com- bine to lead us into a danger which is never far away in the study of dead languages. We suspect that there is a certain semantic difference between two alternative formulations; we find certain examples in which the difference of formulation coincides with this semantic difference; upon these examples we base a general rule; we then * See p. 37. 1, p. 53 mT. > Goodell, pp. 22ff., 27f., 38f., remarks on this danger. I find it impossible to agree with Ebeling’s interpretation (pp. 236£.) of sentences which seem to him to illustrate his general principle, and I find much room for similar dis- agreement, both on principles and on their particular illustrations, with Fischer, pp. 194ff., Holwerda, p. 45, Richter, pp. 33f., Schéne, Umstr. pp. r71f., Meillet, pp. 365 f., Wundt, pp. 350, 368 n. 1. 3 DG34 GREEK WORD ORDER translate all other examples of the alternative formulations in such a way as to make them conform to the rule; and finally we treat our translations as evidence for the validity of the rule." (ii) LOGICAL CATEGORIES Rules defined in terms of ‘logical subject’ and ‘logical predicate’ (or ‘psychological’ or ‘cognitional’ subject and predicate, or ‘determinand’ and ‘determinant’, or ‘thema’ and ‘rhema’)* escape to a large extent the disadvantages of rules defined in terms of emphasis. They avoid the risk of confusion between the rational and the emotional, they leave little room for disagreement on analysis, and still less room for disagreement on questions of degree. The essential difference between ‘logical subject’ and ‘logical predicate’ may be illustrated by taking some simple English utter- ance such as ‘dogs bite’. Irrespective of context, the syntactical relationship between the two elements of this utterance is constant; ‘dogs’ is always the syntactical subject and ‘bite’ is always the syntactical predicate. If the context of this utterance is a discussion of the habits of dogs, syntactical subject and logical subject coincide; ‘dogs’ denotes the subject-matter, 16 Trepi of, Td trroxelyevov. If, on the other hand, the context is a discussion of creatures which bite, the logical classification of the elements of the utterance is the reverse of the syntactical; ‘bite’ becomes the logical subject, and ‘dogs’ the logical predicate. In English, the syntactical categories determine the order of words, the logical categories the volume of the voice. * This criticism may, I think, be levelled against Barbelenet, Etre, pp. 63 ff., 103f., and Ammann’s semantic classification (Wat. 11, pp. 300ff.) of examples of atpety and éAetv in Homer. * ‘Logical’: Weil, pp. 14 al., Holwerda, p. 24, Gardiner, p. 273, Sandmann, pp. 101 ff.; ‘psychological’: Gabelentz, Weit. pp. 129, 335, Sprw. pp. 365, 370, Paul, p. 236, Dittmar, p. 40; ‘cognitional’: Sandmann, pp. 142ff., 245 ff; ‘determin-’: Dittmar, pp. 37ff., Richter, p. 25; ‘rhema’ and ‘thema’: Loepfe, p. 23. Firbas, Comm. p. 39, WO, p. 71, Non-Th. p. 171, distinguishes between ‘theme’ or ‘communicative basis’ and ‘rheme’ or ‘communicative nucleus’. T do not imply that all these authors use all these terms synonymously; I cite them simply as examples of distinctions which are in varying degrees analogous to the grammarian’s distinction between subject and predicate, but must be drawn in the light of the logical relation of an utterance to its context and are not necessarily revealed by grammatical form.LOGICAL DETERMINANTS 35 The difference between the logical categories can be described in several different ways:_-- Gi) In the two contexts which I have postulated, the utterance ‘dogs bite’ is an answer to an implicit question;! in the first context, ‘what do dogs do?’, and in the second, ‘which animals bite?’ In each case the utterance could be reduced to the logical predicate alone; it would be laconic, but it would be intelligible; the logical subject is the element which is common to question and answer.” Gi) The logical subject is what one would leave out if one were sending a telegram the logical predicate is what one would leave in.3 (iii) The logical subject is the element which is expected or pre- dicted by the hearer; the logical predicate is the element which is new, unexpected and unpredictable to him.4 To amplify now the concepts ‘dispensable’ and ‘predictable’: (a) In speaking of a word in a given context as ‘dispensable’ I do not mean necessarily that it could be omitted without any gram- matical adjustment of what remains; thus in saying that émOupsis is logically dispensable in Ar. Nu. 435 Tevfer Tolvuv dv tyelpers: ob yep veyéAav Em@upeis I recognise that its omission would make the change of peyéAwv to weydAa desirable. (6) Words are not dispensable or predictable solely by virtue of their relation to the verbal context; their relation to their material context is also relevant.) In $/G3, 358 a Hidpov & Acivopéveos kal Tol Zupoxdaior To: Al Tupdv’ dd Kuuos, the word ‘dedicated’ is dispensed with because it can be understood from a combination of the words themselves with the nature and location of the object upon which they are inscribed. (c) Whether or not dispensability and predictability are deter- minants of order remains to be seen; but we can say for certain that in any language whatsoever there are circumstances in which order * Weil, p. 22, Gabelentz, Sprw. p. 366, Wundt, p. 349, Dittmar, p. 40, Gardiner, pp. 273f., Hatcher, pp. 239 ff. * Loepfe, p. 28. 3 Cf. Gabelentz, Sprw. p. 366, on exclamations, * Goodell, pp. 30f., Richter, p. 13, Loepfe, p. 25, Bolinger, pp. 1118 ff. 5 That is to say, the whole situation of speaker and hearer, or of writer and reader, must be taken into account, and not merely those aspects of the situation which are put into words. This is stressed by Brugmann, Vschd. pp. 13. Gabelentz, Sprw. p. 370, Richter, pp. 13, 15ff., de Vries, p. 23, Loepfe, pp. 18, 35f., Gardiner, passim. 3236 GREEK WORD ORDER determines predictability. In X. HG, 1, 2, 16 dv fipxe MévavSpos the relative pronoun dv tells us nothing certain about the content of the _ clause which it introduces; Xenophon may, for all the hearer knows, be about to say ‘of whom the majority had lost their shields’; we cannot predict fpye from dv, and still less can we predict MévavSpos from fipxe. On the other hand, in An. 1, 7, 11 &v ’Aptorytpons fipxe, given the words dv *Aptaytpons, the following #ipxe has a high degree of predictability, for after Sv and a man’s name the author is more likely to be going on to say ‘led’ or ‘commanded’ than any- thing else. Similarly in JG, 17, 865B [A]6pos [tené]vos *A[ptenst]&0s *A[uapulalas, Spos does not tell us ‘boundary of what?’, nor does tepévous tell us ‘sacred land of whom?’; hence the second word is not predictable from the first, nor is the third predictable from the first two. But in 865. Adpos “ApteulSos Teyévos *“Avapy[a]ios the word tepévous is both predictable and dispensable, for ‘boundary of Artemis’ can only be ‘boundary of the temple/sacred land of Artemis’ ; compare the complete inscription Adpos Aids (1G, 17, 863), in which ‘temple’ or ‘sacred land’ is understood. (d) The criterion of dispensability in Greek is, or can be made, a very strict one. We are entitled to say that a word in a given utter- ance is dispensable if we can find, preferably, an identical utterance, - or, failing that, a very-similar one, in the same author, or in the same place and period, in which it is actually dispensed with. Thus we can say that in ‘ Callias dedicated me to Athena’ the words ‘dedicated me’ are dispensable because we have actual examples of ‘Callias to Athena’. (e) Dispensability and predictability are not always the same thing. Compare with dv fpye MévavSpos the modern English practice of writing the name of a man in brackets after the name of a ship ora military formation. By convention, the name in brackets is the name of the commander; hence in English the words ‘com- manded by’ are dispensed with. In Greek this convention is not used, so that in dv *Aptoryépons fipye the word fpye has a certain degree of predictability but is not dispensable. Cf) ‘A certain degree’ is an unavoidable qualification. A word can never be wholly predictable from the preceding words; at the best it can only be overwhelmingly likely, and more often it is predictable only in the sense that it is the most likely of a small number of feasible alternatives. Similarly, in speaking in paragraph (d) of ‘identical. . .or. . .very similar’ utterances I tacitly admittedLOGICAL DETERMINANTS 37 that in one sense dispensability also may be a matter of degree.t Nevertheless, dispensabjlity and predictability differ from ‘emphasis’ in that while admitting of degrees in a positive sense they admit of absolutes in a negative sense. I do not know what would be meant by calling a word ‘absolutely emphatic’; but in any context there are words which are ‘absolutely indispensable’ or ‘absolutely un- predictable’. So long, therefore, as certain characteristics can be wholly absent from some of the elements with which we are dealing, the risk of imprecision entailed in the fact that their presence in the remainder is a matter of degree is an acceptable risk. The starting-point of this discussion of dispensability and pre- dictability was the traditional distinction between ‘logical subject” and ‘logical predicate’. This traditional distinction is by no means coincident with the distinction between the dispensable and the indispensable or between the predictable and the unpredictable. The position is rather that when we devise examples of the simplest kind to illustrate the traditional distinction as sharply as we can these examples draw our attention to other ways of describing the logical relation between the components of an utterance and suggest the possibility that these ways may be more fundamental in character, and of a wider application, than the traditional terminology. This possibility is realisable. In postulating two contexts for the English utterance ‘dogs bite’ I chose two of the same type, in which one element is ‘given’ by the situation or ‘inherited’ from what has preceded. This, however, is not the only type of context. Suppose that instead of the habits of dogs or the number of biting animals the context were concerned with the means by which animals defend themselves. In this context ‘dogs’ and ‘bite’ would receive in modern English different intonations but the same volume of voice. If we analyse this utterance in terms of dispensability and predict- ability, the answer is short and clear: both elements are indispensable and both are unpredictable. Yet it is customary to apply to this type of utterance also the analysis into ‘logical subject’ and ‘logical predicate’; and most people familiar with these terms, if they were required to analyse ‘dogs bite’ in the context ‘how do animals defend themselves?’, would say that ‘dogs’ is logical subject and ‘bite’ logical predicate. But would this analysis be meaningful? * Cf. Firbas, Comm. p. 42, WO, pp. 71f., on degrees of ‘communicative dynamism’; but see p. 53 n. 1 below. ‘38 GREEK WORD ORDER In the type of utterance in which one element is ‘given’, the speaker is saying something (the logical predicate) about something _.. else (the logical subject). There are other types of which something similar is true; a list with a heading, or a scholion with a lemma, is an utterance in which something (the list or the scholion) is ‘predi- cated’ of a ‘theme’ (the heading or the lemma). Itis sometimes open to us to conceive a simpler and less formal utterance in this way, and to make our conception clear in our expression, e.g. ‘as for John, they caught him’, or ‘the one that got away was John’, converting, as it were, ‘they caught John’ into a heading ‘John’ and a minimal list ‘they caught him’, or ‘John got away’ into a heading ‘the following got away’ and a minimal list ‘John’. The relation between theme and predicate has played a larger part in discussions of word order than it deserves, for the statement ‘the theme precedes the predicate’ is tautologous, and ‘the predicate precedes the theme’ self-contradictory. One can no more predicate something of a theme not yet expressed than one can ‘contribute to’ a discussion not yet proposed or begun. In the case of English utterances of the types ‘as for x, y’ and ‘the one who x was y’ we have formal linguistic grounds for saying that xis the theme. But where we have only the order x, y to guide us, we cannot infer that the speaker necessarily conceives x as theme; we can only say that y is not the theme.' Most actual utterances have neither theme nor predicate. Dionysius’s story about Plato’s attempts to find a pleasing arrange- ment of R. 3274 KaréBnv xGds els Tepe pete Trcnncovos Tot *Apiotevos is bien trouvé. No element in this sentence imposes itself as the ‘theme’; and however determined we might be to analyse it as saying something about something else, we could never find cogent grounds for deciding whether it says something about yesterday, or about Socrates going to Piraeus, or about Socrates being with Glaucon. It is arguable that attempts to analyse all utterances in terms of theme and predicate rest simply on a logical muddle, to which Aristotle’s metaphors and the ambiguities of the word ‘subject’ in modern European languages have contributed significantly. Where the order SP is a syntactical rule, when the theme of some utterance conceived as theme and predicate happens to be the grammatical 5 ¥ In the English examples the syntactical structure indicates to the hearer or reader the speaker’s conception of the utterance; order by itself does not.LOGICAL DETERMINANTS 39 the normal order satisfies the requirements of the relation between theme and predicate, i.e..74P is expressed as SP. It is, however, an obvious fallacy to argue from ‘ ZAP is sometimes expressed as SP” to ‘xy always expresses 7AP’.* To use the term ‘logical subject’ of the given element in an utterance, of the theme in an utterance conceived by the speaker as theme and predicate, and of the element which a student of the language may choose to regard as a theme, defeats the original purpose of making a distinction between syntactical and logical categories; and it obscures the fundamental difference between dispensable and indispensable elements by applying the same term now to a dispensable element and now to an indispensable one.” Pl. R. 3274 contains five elements no one of which is predictable and no one of which is dispensable in the sense that if it were omitted it could be understood from the context.3 If we are to retain the terms ‘logical subject’ and ‘logical predicate’ we must say that the utter- ance consists simply of five logical predicates; which would be a bizarre use of the word ‘predicate’. For these reasons the tradi- tional terminology seems to me inappropriate to the description of * The consequence of failure to disentangle different modes of classification is reflected in, e.g., Goodell, p. 22, ‘the grammatical subject is likely to be the logical subject’, or Gabelentz, Sprw. p. 369, ‘das Gehérte verhilt sich zu dem weiter Erwarteten wie ein Subjekt zu seinem Pradikat’, or Kieckers, St. Vb. pp. 132ff., who seems almost to equate ‘first element’, ‘theme’ and ‘logical subject’. * This is one of the two major defects (see also n. 3 below) of Loepfe’s analysis, pp. 30, 51. He uses ‘Thema’ of the ‘given’ element, ‘Neues Thema’ of the element which is not ‘given’ but seems to him to have some affinity with the ‘Thema’, Hence in Menander, Epitr. 149 ff. tedtaca1 tpayobous, O18" Sm, Kal tatrra Koréyers mdvra. NnAga tive MeAlav 7” exeivous etipe trpeoPurns dvijp attréAcs he classifies NnAéa. . .éxelvous as ‘Neues Thema’. 3 Paul, p. 236, discussing an utterance of which no element is ‘given’, points out that each element could legitimately be regarded as ‘predicated’ of all that has preceded. Gabelentz, Weit. pp. 136f., discusses a news item in similar terms, but obscures the point by choosing to regard the first element as ‘psychological subject’, of which all the rest constitutes the ‘psychological predicate’. The root of the trouble is the axiom (e.g. Ammann, Dopp. pp. 19f., Loepfe, p. 24) that the majority of utterances have one ‘rhema’ apiece. Loepfe, pp. 37f., therefore introduces the term ‘Nachtrag’ and analyses Pl. Lys. 2034 thus: Eropevépnv piv (Th.) &€ *AxoSnpelos (Th.) eveu Auxelou (Rh.) Thy feo telyous (Ntr.) Um’ odrrd To telyos (Ntr.). Thereby he throws away all the advantages which might have been gained by his other- wise perceptive and original principles of analysis.40 GREEK WORD ORDER the contents of an utterance in terms of their logical relations to their context. I realise that one ought not lightly to impose a new_,. terminology upon a subject already overburdened with termino- logies; but this consideration is outweighed by the obvious disad- vantages of using terms which have already been used in the same connection in different senses.’ I propose, then, to treat a Greek utterance as composed of elements of two logical types, ‘nuclei’ (symbol V) and ‘concomitants’ (symbol C). I call an element V if it is indispensable to the sense of the utterance and cannot be predicted from the preceding elements, and C in so far as it is deficient in either of those qualities.” To describe in terms of NV and C the utterances so far discussed: in the context concerned with the behaviour of dogs, ‘dogs bite’ is CN; in the context concerned with biting animals, it is VC; in the context concerned with defensive habits of animals, it is WN; and the opening sentence of the Republic is NN pN pN pN. There is one further type of utterance which must be considered. Imagine someone saying: ‘Keep clear of dogs; dogs bite.’ If this utterance were written down, without the help of italics or under- lining, any reader would interpret the second occurrence of ‘dogs’ as C. It could logically be replaced by ‘they’ (pronounced with diminished voice) or, indeed, omitted without serious loss of intelligibility. But the fact remains that when an utterance of this kind is spoken both elements can receive equal volume of voice. In other words, the speaker is treating ‘dogs bite’ as if the words belonged to a context different in type from that to which they actually belong. If I ask myself in what circumstances I would do this, I can give a definite answer; I would do it when I had in mind a contrast between dogs and other things; I should be implying ‘other things may not bite, but dogs do’. Implicit antithesis would make me treat as Wa word which it was open to me to treat as C. Now it is one thing to explain what one would have in mind if one intoned and stressed in a certain way some words of one’s own language; it is quite another thing to demonstrate what was in the * Cf. Richter, p. 11, on the desirability of a new terminology even in 1920. Sometimes the same term has been used in opposite senses; e.g. Ammann’s “Satzbasis’ (Unt. 1, p. 13) is the opposite of Firbas’s ‘communicative basis’. ? *Nucleus’ is used by Firbas, see p. 34 n. 2 above. ‘Nucleus’, ‘satellite’ and ‘concomitant’ are used by Pittman, pp. 288f., with phonemic and morphemic connotations.LOGICAL DETERMINANTS 4I mind of an ancient writer.’ If I formulate rules of order in terms of Wand Cand then explain away every recalcitrant example by saying that the writer must have chosen to treat C as NV, I am wasting time on a grossly circular argument. Whether or not the situation is as difficult as this can only be discovered by seeing whether there is a ‘normal’ pattern of relationship between NV and C and how far the abnormal instances can be defined and classified. B. ConcoMITANTS . (i) TREATMENT OF CONCOMITANTS AS POSTPOSITIVES By definition, M and q are such that when an utterance is composed of one of each they are arranged in the order Mg. We have seen that inearly Greek the combination of two M with one or moreg tended to be arranged in the order Mg(z) M. Can we tracea similar process in logical terms, of such a kind that the combination of N with C takes the form WV C,? and of two N with C the form N CN? Again, we have seen the effect of a tendency in historical times to treat a clause as a pair or series of word-groups and to distribute g among these groups. Is there a parallel logical development resulting in the distribution of C over a series of word-groups each of which begins with V3 The pair of documents which follow were inscribed at Tegea in the fifth century B.c. (/G, v, ii, 159). (B) was intended to replace and cancel (A), but fortunately (A) was not cancelled with such vigour as to render it illegible. (A) @) §1. ZovOlar tS Orroyadd Zoudlar mapkadéxa 731 Oiraryats Bixécrica pot TRETPOKETICN HVAT doyuPIS * Brugmann, Vschd. pp. 4f., emphasises the essential difference between the study of spoken and of dead languages. 2 Suggested (in different terminology) by Ammann, Unt. 1, pp. 20, 26£., Delbritck, Altind. pp. 51f 3 The argument from analogy is in itself of limited value. The fact that in most Greek words the terminations give the spatio-temporal orientation of the stem accords with the arrangement Mg, where g gives similar orientation of M. Yet attempts to infer general ‘ Bestimmungsgesetze’ from the structure of the word and to apply these to the structure of the sentence (e.g. Dittmar, pp. 37ff., Goodell, pp. 21, 34, Hirt, p. 235, Gabelentz, Sprw. p. 373, Bergaigne, Pp. 22ff., 125 ff.) do not do justice to the facts. .42 GREEK WORD ORDER The words trapxa@rKa and dpyuplo are C, because they could have been omitted and understood from the material context; the demon- stration of this is provided by (A), 1, where the composer thought it unnecessary to say that the money was a deposit or that it was in coined silver. The logical pattern of (B), 1 is: WC pNiNNC; 1 take tTzetpoxérici uvat as a complex V of numeral type. Cf. JG, 1x, i, 333.4 (Locris, V z.c.), 4ff. af x’ &8{K5(s) ovAd1, tTétopes Spoyuad al 88 TAgov 5&’ duapav Exo1 Td aUAO, AEuISAlov SpAETS FoTI CUAGCA, DGE, 412 (Olympia, VI B.c.), 1 at 5& Bevéor tv tlapdi, Bot Ko G6c501 kal Ko8épo1 TeAcicn. In the first provision of the Locrian document, the composer did not think it necessary to say that ‘four drachmae’” was a fine, given that the document is a law and laws are largely concerned with the specification of penalties. Therefore épAéta in the second provision and 048801 in the Olympian document may be regarded as C; they exemplify respectively the logical patterns NCpN and Ng CpN N+ The remainder of the Xuthias documents from Tegea is: § 2. af K’ attds Aixé, el pév Kec 368, eveAtods. onrrds cveAto8S. §3. al 88K’ drrobdval, ail BE Kor pe 368, Tov Thkvov Euev, tol viol éveAda5 Tol yvEcio1, brrel Ko trévte Fe tec AEBSveT1. Entel Ka EBdoovrr TrevTe FeTECt, §3a. el 8E Kor pé 3501, Tad Guyorrépes dveSa86 Tal yveota, §3b. al BE xox pe 35vm1, Tol vdbor éveAdo06. § 4. at BE Kor pe yeved Aeitréron, ef SE Kor pE voor zon, tv EmBixcrov Eyev. Tol *5 Goro THébiKEs cveAda8d. §s. el 5é x’ dvpidtyovm, BSiayvopev 88 tds Teyettas tol Teyedron Sixyvovrd K& Tov GeBpdv. K& Tov Gebpdv. In (A), 2 dveAéa8eo is N; it is, in fact, a complete clause in itself. In (B), 2, on the other hand, the W status of éveAgoOe is in peril; given ovrrés, the sense of dveAéo6o becomes almost inevitable, and the * T do not distinguish between NV, CN, where N, and N, are syntactically interrelated (e.g. as S to O or as components of S) and the same arrangement where N, and N, are co-ordinated (e.g. S, 7+5,). Cf. Havers, Spalt., Hab. pp. 44f., Boldt, pp. 78ff., 103 ff. Rass, Krause, pp. 245 ff, Delbriick, Altind. pp. 58f.LOGICAL DETERMINANTS 43 word might therefore be regarded as dispensable because predictable. The composer has chosen, by means of the order, to give C status toa word not inevitably of that status. But in (B), 3-4 the composer no longer has a choice; given éveAéo6e in (B), 2, dveAcso6co in (B), 3, (B), 3a, (B), 3b and (B), 4 is necessarily C. When combined with a single N, in 3b, it follows it; combined with two J, in 3 and 3a, it is placed after the first N. In (B), 4 I take &ciota TréOiKes to be a complex N, so that the pattern is ppNN C. A more complicated and sophisticated document, the First Decree of Callias (Athenian Tribute Lists, p 1) illustrates the same principles: (Gx) err0BSva1 Tois Geois Ta Xpéycrra Ta SpEAS- NpN pN pN...|| (§3) Aoyiodabav S Aor AoyioTal. . . Nq pN...|| (§4) ouvayoyés 58 TSv AoyioTSv Nq pC pN N M!|| é Bont osrroxpérp ford. (§5) érrobévrav Bi-1é xpsuorra Aor mpuréves... Mg pC pN...| géréoovres Th Te Trivénict Kol Te ypon- Mpg ppl... .|| berteicn . (§6) cerogaivévrdy 8 ta yeypoupéva hol te Ny pC pglV...|| Auepés.. . In§4 t&v AoyioTév is C by virtue of of Aoyiotal in §3; in §5, Tx Xprpora is C by virtue of §x and (by implication) §§ 2-4; in § 6, ta yeypanuéve is C by virtue of the latter part of §5. The decree goes on to provide for the creation of a new board of treasurers; it defines their essential function, and then proceeds to details: (§9) Tope 88 tév viv Tomdv Kod Tév Eriota- pop C ppN.. .: cmrapiOpéodo86v Kal érrooticdobay ta NpN pC...|| XPEnOTA. .. (§10) Kal rapadexado0Sv Aoi Tapioca hoi Aaxdv- pM pC pN ppN C|| Tes Tapa TSv viv d&pxdvTov With érropitpnedotey Kal crootnodoteoy Ta Xpriporra in §g cf. & Bopéns te kal $ yetpdov éot&or in the following passage of Herodotus (i, 26, 2): el 88 f oTdo1s fMoxto tev copécov papN N pN|\ kal Tod olpavol * ppN:i TH] Lev viv 6 Bopéns Te kai 6 xeicov Eot&o1, pgNipNg ppN C\i44 GREEK WORD ORDER tout uty Tod vétou Av t otdors Kal THs MgipN Mt pC pph'|| LecauBpins, Tf} BE & vdTos viv EoTnKe, PapNiN C\| ‘Towrtn Bt 6 Bopéns, NgipN'|| al tattra obtws elye, pPNiN C|l 6 HAlos dv PNG érreAcuvdyevos eK udcou Tot ovpavot NpN pN: Ud TO XEIpdsvos Kal TOU Pops pPN ppN| fie dv te Gveo THs Evpcotrns Nq pN pN\| xorrécrep vGv Tis AiBins Zpxetan, PNipN C|| Siebidvra 8 &v piv Sic tr&ons Evipeotens Naqq pN N: EATropcat N troiteiv &y Tov “lotpov Nq pN\| técrep viv épydczeton Tov NeTAov. PN CipN Certain words in this passage are immediately identifiable as C: kotGan, (fv) 4 otéors, and foTxKe, by virtue of the initial 4 ordors; &pxeton because of fie dv, and épydzeton because of trove &v; and elye on grounds of dispensability. One of the commonest brachy- logies in Greek is the omission, in a relative clause, particularly with rep, of a word intelligible from its occurrence in the preceding clause; Herodotus here, as often, avoids brachylogy by expressing words which could be left implicit. 6 Bopéns Te Kod 5 yeiev are a pair of N co-ordinated; so are to véTou Kal THis uecouBpins. In the former case the C is placed after the second N, in the latter case after the first. Towards the end of the passage, viv and ris Aipims are V; so are viv and tov NefAov; in the former case the C tpyeren is placed after the second N, in the latter case the C pyagetan after the first N. Tn this short passage we have four clear examples of C combined with two NV. In two cases the W are co-ordinated; in the other two cases they are not. Whether they are co-ordinated or not, the C may be placed either after the first WV or after the second. Platonic argument illustrates the same principles as legal docu- ments and Herodotean exposition. In Laches, 194& Nicias expresses . the view that courage is Sewésv Kal GappaAgov émotiun. After dealing with a certain amount of obstreperousness from Laches, Socrates embarks (196D) on an examination of Nicias’s hypothesis, and begins with a formal statement of it: Thy év8pelav émothynv * Thave provisionally treated £Atropen as NV, but see (iii) below. 2 See p. 42 n. 1 and v below.LOGICAL DETERMINANTS 45 pis Seedy Te Kal Gapportov elven: pV N Mt Ng pN M?, In 198D he turns the argument tathe discussion of tmorfjuen in general, and in 1998 begins his conclusion ovKotv, dpiote, Kad t dvEpela Téav Sevésv Emorhpn totlv Kal Gappadtov: N...ppN pN CM: pN. He reminds Nicias what té Swe Kol te Gapporéa are; he reminds him of the nature of &moriyn. Then: ob udvov dpa té&v Savev kal Gapparscov 1 dvbpela Emorihyn tory: W Mg pW pN pN C M+. The argument continues (199): ‘You have told us about a part of courage; but we were asking what courage as a whole was. Now it seems o¥ pdvov Savesv te Kal CappaAgov émotiypn i) dvbpela totiv’: NNN qpNN pC M, My analysis here does not depend upon subjective interpretation of ‘emphasis’, but upon the simple test of dispensability. In 1998 ov pdvov &pa, KTA., given the preceding few lines, | dvBpeter could not be omitted without causing confusion, but &moripn éoriv could. In r99¢ the reverse is the case. With the exception of elye in Herodotus’s ef tatrra otrress elye, all the words which I have so far treated as C have been words ‘given’ by the verbal or material context. There is in addition a common type of C (elye is an example) which is dispensable in so far as the word-group which contains it could be rephrased to convey the essential meaning without it. An easily applied test is to substitute q or M*. Thus in Pl. Grg. 470D EvSaipcov ob oor Soxel elven 4 SPAi05; OtK olBa, & Madde, oF yép tra ouyytyova TH dvbpl, the words 16 évBpi are pC because they could be replaced by the g avré. Cf. xetron in two passages of Herodotus: VII, 198, 2 (. . .GAAos TraTods) TH otvoucr Kettoa AUpos pV CN = olvopd tot: Avpas pPNMN =Té obvopa Aupas PNN 200, 2 (Kdpn Te EoTl) TH Obvopa “AveHAn Keira = pV NC =rij owouat “AveriAn fori pN N Mt =i} otwouct “AVOFAN pNN In cases to which the test of substitution is inapplicable there is room for much doubt and disagreement, according to one’s view of the extent to which the sense conveyed by the omission of the alleged C falls short of the ‘essential’ sense. For example in Pl. Euthyphro, 24 ros tv Auxel xorroArmav Siorpipds évéaSe viv Sicrrpifers there is antithesis between &v Auxelco and évé&5e and be- tween KoroAitre and viv; these are certainly N, and Sicrrpipers is46 GREEK WORD ORDER certainly C; but I may not. command agreement in suggesting that BiorrpiPass is also C (res tv Auxeteo. . .SiartpiBés being replaceable by 1d Avxsiov), so that the logical structure of the whole utterance is PPNINC|NiNC. Cf. D. xiv, 3 fuels 8 domrep tv068" elcoGerpev, otra Bryony Kal Ewo=M(=M*) gipN C||M(=M*) CipN, in which eldGeuev might plausibly be regarded as WV; it is its sub- ordination to the strong antithesis évO&5e/ffco which gives it its flavour of C. The passages already analysed have shown the analogy between Mi Mg and NN C, and I add one further example of the ‘distribu- tion’ of C, Lys. 1, 15-16: mrpootpyeral pol is TrpesBOnis bvEpestros qq NC| Urrd yuvoukds UTroTreppbelon. . . PNN...|| oti BE (se. yuvit).. M=M*)q...| TrpodeABotion: ot por Naqi bys f EvOpeortos THis olklos THis euiis emm- pC pN pN | “EvoiAnte’, épn,.... NC... Here the C 1 &vOpcrros is placed after the leading word of the second word-group of a sentence in the manner of e.g. the g pot in D. xxt, 26 obvavrios Fxev dv elOis yor Adyos. Gi) CONCOMITANT GROUPS Naturally the C element in a clause may be more than a single word. The simplest type of complex C, like the simplest of complex , is a pair of words which together constitute a familiar expression and are rarely separated, but a clause may also contain a succession of C which all have similar logical relations to the preceding clause(s) but are not united by any other common factor. Both types are illustrated by §§ 10-12 of the First Decree of Callias: (S10) ...xod ev oTeAd dverypagedvtév §=ppNNNN...|| wen Grrowrar. . . (S11) Kal 1d Aortdy dvaypagdvrdv hor = ppNC pNC pC|| ciel ropion bs oTEAEV . kad Adyou SiSévréy t3v te bvTév «= pN N pg C ppNC p|| xpewérév Kal tév trpocidvTéy tols Geois. . : ”LOGICAL DETERMINANTS 47 (§12) Kod & Towatévalsy &s Tavativare pp pN pCC TOA Adyov BiGdvrév... In §12 & TMavatnvotev és Tovoéhveia is a complex V of familiar type; Tov Adyov BiSdvreov is a comparable type of C. In§11 the V are 7 Aormév and of atlel ; roqufoa and &5 orHAnv are both C by virtue of what has preceded, but the C status of each of them is independent of that of the other and of éverypapévrawv. Three literary passages illustrate the treatment of C groups; they are arranged in ascending order of magnitude: (a) Euthyphro, 82 arpdfecks Tivos trépi Siexpepdpevor Noq N| of pay BSixadoos parol ory Trempéx6on, NqiN Cq C|| of 8 GBikcos NgiN Here paciv is strictly speaking M* in character, and the word- group which I have analysed as N Cq C is therefore on the borderline of the category ‘C group’; it admits of the analysis NMC. (6) Chrm. 164D . Kod ouppépoyian TE av Acrgols dvadives pV ppN N pN N|j 7d Todo ypduyc. Kol ydép totrro ofte yor Soxel Th poi Ng CpCC|p... yedupa évoxsioter, as... The word-group headed by the WV otros stands essentially in the same relation to Toto as may be seen in the much simpler Hero- dotean clause el totrra! otrreas elye. The first C of the group, Soxel, is, like gaotv, a word of special status (see section (iii)). (c) Hdt. 1, 1, 1 (...T& Te GAAa Kad 81’ Av altiny éroAguncav AAA AoIn.) Tlepotcov pév vuv ol Adyior Neqq pN: Polvixas attious pal yevtobar tis Siarpopis «NCCC pC Here again gai and yevéodan are words of special status, classifiable as Mr‘, so that the word-group Qolvixas...diapopfis may be analysed as N(=M?) C(=M) C(=M®) C(=M®) pC(=M), and is clearly modelled upon Mi Mgq M.48 GREEK WORD ORDER ii) TREATMENT OF CONCOMITANTS AS PREPOSITIVES _ The analogy between Mg M and N C N suggests the possibility of a similar analogy between pM and CN. There are, I believe, certain categories of word which are commonly treated as if they were p. Consider, for example, Hdt. 11, 81, 2: katrot Tupawou UBpiv petryovtas dvSpas PINNNC: 8s Styou dxoAdotou UPpiv trecely pNNCNG?) tom obSapdds dvaoyetdv. ton N Nl Whiat is the status of tom: here? Is it, as M', to be compared to q immediately following a word-group which it is nowadays customary to mark off by commas (ch. uy, (i), (D, p. 13)? Or is it wrong to suppose that there is any kind of pause in the voice after Tecelv? The latter question, at least, cannot be asked about III, 82, 1: dyol 88 r& pv fre Mey&Buzos Nq\pq ene V: & 1d TAABos Exovrat PPN C|| Boxdet otk Speéds Aéfan. Boxér WV C|| Both ele and Soxée are obviously C in character, since they are logically dispensable; yet both of them here are the first M of a clause. Cf. also Hp. Carn. 4, 3: ef Tis SEA01 Srrrév pq Bt N Ng pN pNy|| veupcaBedk re Karl KoAACBect karl TEAL 5, ‘ret pay 6a Tay Srrtéron, paNiN C|| Te BE veupcdBed Te Karl KOAACBEct pyNg pNINCC...\\ ou bkAst drrtGobat... 1 BE tridtarrov Kad Arttapov PIN pN:N C|| tégiota Strré&ron, The passage exemplifies the same logical principles as that of Herodotus on the path of the sun, with the single exception of 8A01; cf. ibid. 19, 6 ef BE TIs BoWAeTan Kad Toto éAfy Eon, «TA., where PowAeten is as insignificant for the essential sense of the passage as £01 in 4, 3. Tt would therefore seem possible that words meaning ‘think’, ‘seem’, ‘want’, ‘be willing’, ‘say’, and ‘be’, when there is no significant antithesis between thinking and saying, wishing andLOGICAL DETERMINANTS 49 doing, being and becoming, or appearance and reality, can be treated as if they were a This is by no means the full story of the behaviour of elven (see section (iv)), but it offers an explanation of the very common type of clause voplzeov otreas dv Gpiora Trpa Eon, in which g is placed not after the word meaning ‘thinking’, ‘knowing’, ‘hoping’, etc., but after the next M. It is noteworthy that these words are dispensable in the sense that they could be replaced by os with a participial construction or simply by an accusative and infinitive construction.? (iv) PREFERENTIAL TREATMENT OF CONCOMITANTS Demonstrative words are often logically C, but they are exempt from the treatment which I have described in sections (i) and (ii); their role as M* may take precedence over their logical category. Thus we find in the First Decree of Callias: (G7) Teaulos 8 érroxvapevey Tourrov Tov NgNCpC...|l (§8) Aotrron St toyievovrév ty dA... M7, CpN...|| otro1 in §8 is as dispensable, logically speaking, as tovrraw in§7, but is nevertheless given precedence. This treatment of demonstrative words is extended to words which are in explicit antithesis;3 the nature of the extension may be seen in many passages of Herodotus’s catalogue of Xerxes’ army, e.g.: Vl, 62, 2-63 Kioioi 88. . .urtpngdpor foav. Nq...N Mil Kiootoov 88 fipye ’Avdons. . . Cy NN...| *Aoaupion &. . .tyyeiplia. . .elxov.. . Nq...N...C...|l otro: BE td uty ‘EAAtvav exchtovro §=— MgipgN NN...|| Zpior... fipxe 88 agecov "OTd&otms. Nqq N\| * Firbas, Comm. pp. 45f. * Fraenkel, Kolon, pp. 327ff., treats voulzov, thyotpevos, etc. in such utterances as ‘Kurzkola’. 3 Demonstratives often imply antithesis, but their treatment as M* does not depend on their antithetical element; cf. o¥ 58=M’q in cases where antithesis is ruled out, e.g. Hdt. 11, 69, 3 Kod fv uty patuntot Exov dra, véurze cecouTiyy Zuepbi Te Kipou ouvorxéeiv, Av SE yt Exoav, ob 88 7H udryoo Tpépb. 4 Dego GREEK WORD ORDER 73-4 1 Opuryes 68 éyyorécre THis Maghayountis NgiWpNCC...|| oxeuty elyov... of 88 Opuryes. . .éxaAéovro Bolyes. .. pyc... .NN...|l *Apuévior Bt Kerr: rep Dpiryes toeatyato... NgipgNC...|| Torey cuvaporéowy fipye’Apréyuns... MeN NN...|| AvBol BE d&yxorecros Tis “EAAnviktis elyov Nei N pNC C|| ol BE AuSoi Mnloves exaAeOvro 78 én... pgCiN'C pN...|| From the point of view of dispensability, there is no difference between the ¢ agecov in fipxe 5¢ apecov and the demonstratives oot and Totrov in offtor. . .dxoAetvro and Totrraw. . .fipxe, but the writer has chosen to use demonstratives in the two latter cases; he could have chosen otherwise. Equally, thereis no logical difference between opeav and Kisolav, or between of 88 Moves or of & AvSoi and the third person plural termination of the verb; but the writer has chosen to cast his account of the Persian contingents in the form of a list in which each item is explicitly contrasted with what has preceded and what is to follow. ‘Catalogue style’, in which the initial word of an item may be a demonstrative or treated as a demonstrative, is the most obvious type of extended explicit antithesis.’ A similar preferential treatment of C appears in shorter antitheses, e.g. Hdt. 11, 22, 1: Aeyel yep otB" xin obey, C=p?) q pN N| goueon tov Nefhov Aéew dard trHKopevns — C(=p) pN N pN N| Xt6v0s, &s Abs? pty ex Aipins Sik udowov Alfiérranv, pCzipN pN N'|| &xBiBot 88 és Atyurtrrov. NqipN|| Related to antithesis is the preferential treatment of a word in rejecting someone else’s argument, e.g. Hdt. 11, 20, 2-3: tév t érépn piv Adyar ppNgq Ci tous Ernolas évépous elvor altious PNN Mi CiC pC TrAnBUelv Tov TroTeYOV. . . * Th. v1, 43 is a good example of ‘catalogue’ style. ? It must, however, be observed that there is a logical affinity between fal, in describing the course of a river, and the usage of fon discussed below (p. 52); cf. Kieckers, Se. V2. pp. 58f. 3 This is on—or over—the boundary of emotive utterance; cf. the scornful tepetition in Ar. Lys. 430 ff. ndtv expoyAsvere. . ti Sel poxAdiv; ob yap HoyAGiv Sel, «TA.LOGICAL DETERMINANTS 5r TroAAdaas 8 érnoton piv otk dv Ervevooy, NqiCqi.Ng C\| 4 88 Nethos rootrrd Epydigeres, PAIN:N C\j trpds & ef Ernoin ation Acav, NqipC CM || XpFlv Kad Tos dAAous TroTapOUs, C(=p) ppN C|| S001 Tolan Ernoina dvrio: béouc, ppCNC|\i dpoless tréoyei Kal Koer& wroir& te NelAco. N CppN pN|| The degree of implicit antithesis involved in such an argument is almost the furthest that we can trace the influence, via explicit anti- thesis and ‘catalogue style’, of the preferential treatment of demon- stratives. We are already passing into the sphere of influence of a different phenomenon, the preferential treatment of words which have some emotional force—among which I would class the Aé ToAUs and tras, in which there is necessarily a degree of emotion absent, for example, from &vioi and uépos. A different category: of preferential C is indicated by the relation between §1 and §2 of the First Decree of Callias: (Gx) dwoBSven -rols Oeols ta xphuora... MpN pN...|| (§2) dtroBiSéven 88 dard tSv xpeudrav & CyppC|| ppN M*pCN|| ts dardSoctv tori tols Geols tpce- groytver The explanation of this phenomenon is as follows. It often happens that in a compound word one element is N and the other element C; this is true of ov8els, otxén, etc., e.g. in Hdt. 11, 119, 6: évip uty &y yor EAAos yévorto, N4qqqiN C\|pN NI el Saduoov 26201, kal téxva GAAG, ef Tatra éroPéAom, PN Cipv N |i Trorrpds 62 Kal untpds otxén usu godvreav =. Ng pNENg N| aBeAgeds dv GAAos ovBevl temp yivorto. Ny CINCC|| It is true also of compound verbs, as in D. xx1, 32: ay pay Tolvuy IBicymy Svat’ anrrdv OBpion. —pggV Mtgg No|l TIS. ee ypaphy UPpecos'. . .pevferon, NN...C|| dav 88 Ceopodtrny, Eripos orton Kodderrer§. PIN ||N M N\| Bid 1; STi Tous vduOUs HEN PN \|ppN N 6 ToUTo Trotiv trpcouPpize1. pPNCN * OPpews is not, I think, C, for ypopiv OPpews is balanced by Sixny Koxnyoplas. +52 GREEK WORD ORDER —where mpoovBplzei is analysable into N mpoo- and C -vBplze. A fortiori, a preposition in a complex of preposition with noun or -.- pronoun may be N; and so it is in dard tév xpnudrev above. But if a preposition or an element in a compound word may have N status, so may a termination or any form which belongs to a system of forms. For instance in D. xx1, 122 tls dyola ti Tovrou ytyov’ 4 yévort’ &v trovnpia; the V element in yévorr’ dv is not the stem yev- but the mood index with the g &v. Here, and in expres- sions such as Eom Kod Loon, 8Sdxe1 Kal Soxel, etc., the meaning is indicated in English (but not in all modern European languages) by diminished volume of voice on the repetition of the verb stem and augmented volume on the auxiliary, can, would, did, etc. English ‘emphasis’ may reliably be employed in translating passages such as Hdt. 111, 64, 2 pode 88 dog uderny derodcorsKens ef Tov BeAgedv, Gracie ZpépSiv- dcrroxAcoos Si Kal repinuectioas ti dréon cuLgopi, cvabpdoke él tov frrov...xal of dvadpqoxovn trl Tov Trrtrov TOU KOAEOU TOU Elpeos & wHKns drroreitrrer, étroKAauoars SE virtually=dra, and dvatp@oxovn virtually=tv tovre, requiring the translations ‘after he had wept...’ and ‘wAéle he was jumping up...’, and exhibiting a certain affinity between this class of preferential C and the preferential demonstratives.' The class, how- ever, includes phenomena which are not demonstrative in character and cannot be translated by English demonstratives. The verb elven is sometimes used by Herodotus as a signal that an item of general rather than historical information is being given, e.g. 1v, 158, 2Kal Tov KeAAICTOV TOY Ycopao Tver. . . ut) orev, . . .vuxTds Trapijyov. fon 88 TH xcop~p Tolrrp obvoya “Ipaca, ctr. ibid, 182, 1 KOAcvds TE dads ton Spots TH *Appcovicn Kal Gp, Kal &vOpcotroi rep! ovrrdv olkdouar: Te BE xcopep Tore ofvor AtyiAé font. In both these cases alike'the English-speaker would increase the volume of voice on ‘name’ (‘the name of this place. . .”) and not on any word which could be regarded as a translation of tot? érroBibdven in the Decree of Callias belongs to this category. The aorist infinitive érro8otven in § 1 prescribes the act; the imperfective érro8i8dvor is the signal that a detail about the act is being pre- scribed, and the next words érd Té&v ypnucrov, where dd is NV, show that it is the source of the money which is being prescribed. Cf. Firbas, Comm. p. 42 n. 25. 2 Bloch, pp. 243, Wackernagel, Dicht. pp. 18 ff.LOGICAL DETERMINANTS 53 Cf. IG, 1, 16 (Athens, 3948.c.), 10ff. dydao[1 52 AGnvaricov pav] Tos ospomydfs...Kal tds tiriréas, ’Epetpiteov 88 [ds otparenyds Kal] thy Boat... .[... wal tds GAGs] dyads... duvdvar & Tov vouip[ov Spxov éxorrépo]s Tv Tape aplorw orrols. C. Nuciei (i) SERIES So far we ve have been concerned with the relation between W and C. A less tractable problem is raised by clauses which contain two or more VV. Where both W are M*, what determines their order? To speak of an order of importance is unhelpful, since it is rarely possible to form any opinion of the comparative importance of the different N in the same clause." Certain Attic boundary-inscriptions of late fifth-century date are illuminating: IG, 1, 897 Sep" ['EAsjuowidv [rp]ircrus TeA[e}ursa, Tlepaidy 5 Tprrrds Gpyerat. Ibid. 899 [8]eGe" *Enra[k]péSv tprrris teAcutéi, Opiacidv 8 Gpxeran tprrrus. SEG, x, 374 Jrfprevis *]eA)elurda, Taapparftev 8] dpxer[n Tlorrris. IG, 1, 898 [S]eGpe Mafijoviéy tprerus reAeurén, Gpyetor 58 Muppi- vooly ‘tprrris. : The first member of the antithesis is arranged uniformly in all cases; the name of the trittys, followed by the word tprrtis, followed by teAeuT%. The structure of the second member varies. The two V in it are the name of the trittys and &pyeton, and these may occur in either of their two possible orders. tprrrus is here C, and may be attached to either NV, giving the two patterns WC N and NN C. It thus appears that when two members of the same syntactical structure and similar content are in antithesis, the second member may or may not be arranged in the same order as the first. When it * Firbas, Comm. p. 42 insists on the need to discover the relative importance of all the elements of an utterance, and speaks (Non-Th. pp. 171f.) of a ‘transitional status’ between ‘rheme’ and ‘theme’. I do not feel able to do more than point out, in a given example, which elements have some degree of C status, and would prefer to leave all other questions of ‘importance’ alone.54 GREEK WORD ORDER is not, the order of the antithetical pair as a whole is given the technical term ‘chiasmus’, and may be symbolised (x,y) (y/ x’). The fact that boundary-stones may be chiastic shows that chiasmus is not necessarily a literary embellishment.’ Cf. JG, x1, v, 93 (Lulis, V 3.c.), 14ff. é&rropalvev tiv olkinv eAciGepov Gordoon Tpdtov, Errata 88 towtrot olkétny tuBdvra, DGE, 179 (Gortyn, V 3.c.), off. peBé re 7s yuvannds Tov &vBpa drroSsGan pS" Emomévacn, yé5* uly T& Tes porrpds. Less conspicuous than chiasmus, but related to it, is the series of the type (xy)+(x'y')+(y"x")+(y"x*) ..., in which the two mem- bers of each pair are arranged in the same order, but each pair may differ in order from the preceding pair. The following passage is taken from one of the more artless speeches of the Demosthenic corpus, [D.] xivu, 6f.: iy 8 tye pty HEleooa trapaAcppdvelv, Gedqnuos &8 trpotkarécate Tapasotva, es otrrol pact, 7d 88 ody’ ovSels elSe Trapdv,.. . tuaptupnoay 8” of udptupes otto as EAOI TrapaBotiven Gedqnyos kad TrpdkAnoiv TrpoKaAoITo, Hono 8 of Bixaotal éAnOF elvan Thy poptupiav, garyetv 8° eye tov EAcyyov..., Tras otk dveryKaléy toniv TouTous Tous udpTupas TH wedi] WeHOpTUpNKived ; In the first antithesis (Bye piv, KTA.) Gedpnuos 5é, TpotKaAsoaro and strapaotver are placed in the second member in positions corre- sponding to tye pv, tla and trapaAopBdverv in the first; in the third member 1 8 ope corresponds to ty pév and Gedqnyos Sé in position and in morphological category, though its syntactical role is different. In the second antithesis (QuaptUpnoav 8, KTA.), the verb @r\@noav 8é occupies the same position in the second member as the verb guoprupnooy &é in the first, and the subject-noun of Sixcotal the same as the subject-noun ol ukptupes otto. In the third antithesis (6An@fj, KTA.) the predicates éAnéij elven and getryenv &E take first place in their respective members. * Schick, pp. 370f., Leumann, pp. 797f., Delbriick, Altind. pp. 6rf.LOGICAL DETERMINANTS 55 The fact that one cannot predict when a Greek will employ chiasmus in antithesis is # particular case of a more general pheno- menon: one cannot predict at what point, in a series of members of similar or identical structure and content, the internal arrangement of the members will be changed. There is a large class of utterances which is not as a rule mentioned in discussions of word order but is nevertheless highly relevant to this matter of series; I mean lists, especially lists of payments. For example, each item in the first Athenian Tribute-List conveys two pieces of information: the name of the city which has paid, and the écropyt, of the amount which it has paid. Throughout the list, the name of the city is placed first and the amount second. In the next list, however, the order is reversed, and it remains so in all the other extant lists. The relevance of facts of this kind to the multinuclear clause may appear from consideration of some passages (S/G3, 241) from the accounts of the commissioners in charge of the reconstruction of the temple at Delphi in the fourth century. The entries made in the first part of the accounts for 356 (SIG3, 241, 4ff.), under the rubric 5cxe & 1éM1s, are of'a pattern which is:normal throughout this series of documents. The name of the recipient is given first; then the goods or services for which the payment was made; then the amount of the payment, e.g. Maslov loxeydou pvds Séar. . .“AppoSicn xoAnel Seopdsy pvéts €6. In the latter part of the accounts of 356, however, there is a list of entries in which the order is changed, the goods or services being specified first and the recipient second, e.g. poxevesperros XoupdAcd pds TETopas: BoAlou elopopas Spaypal tpeis, ApioPéAiov: Tot Td poycveoe Alecv Topas Geoyéva Bpayyed trévre, «tA. The rubric which introduces this series is &rrd tovrrou évéAcoua. Yet the change in order can hardly be related to the change in rubric, for the composer seems not to have realised at first the grammatical consequences of the new tubric; he wrote pvas tétopas in the accusative, as if under the rubric Scone & TréAis, and only in the second item passed to the appropriate nominative (a confusion which occurs elsewhere in this series of documents). The abnormality of the order is in fact matched by the abnormality of the circumstances in which these payments were made. The document itself states: per& Tov Aoyiopsv, Topedvtwv Téw BoureuTav, émérafav tol vooroiol. . .dpyupiov Sdyev trot] re Eoya TH Ev KopivOcn. . .Kepcrapa tot FAaBov per&56 GREEK WORD ORDER tov Aoyiaydv of Kopivbior vaotroiol Kai & Tixucdvios vas SexaroKred, The explanation, I think, is this. The draft record of the expenditure __. at Corinth was presumably made not by the man who drew up the record of expenditure at Delphi, but by the Corinthian commis- sioners. Their arrangement differed from that adopted by the recorder of expenditure at Delphi; given a free choice whether to name the recipient before the service or the service before the recipient, the Delphian made one choice and the Corinthians made another. When the Delphian came to incorporate the Corinthian draft record into the final record, although the arrangement differed from his own, he did not think it worth-while to alter it into conformity with his own. The important aspects of this case are: first, that in utterances of exactly similar nature and identical structure, the NV can be differently arranged by different individuals; secondly, that one individual may adhere so consistently to one of two or more alternative arrange- ments that he gives it the status of a formula; and thirdly, that the formulae adopted by different individuals are nevertheless equivalent in the sense that an individual does not regard someone else’s formula as incorrect. Gi) FORMULAE Utterances which contain a S~V’ or O-V relationship are no less susceptible of conversion into formulae than those which do not. Compare the formulae of the prescript of an Attic decree with the equivalent formulae of Argive decrees:* (1) Boke tH Bova Kod te Shep CArodg ESo€e teri (2) 6 Selva Eypoppérreve yrogels Pwrds & Selva: (3) 6 Selva Errecrtérer épptteve 6 Selva (4) 6 Setver eltre Baek 6 Selvar The two states differ radically and consistently. In (1), Athens says ‘decided by the Assembly’, Argos ‘by the Assembly decided’; in (2}(4), Athens puts S before V, Argos V before S. Probably in each state the three formulae (2), (3) and (4) influenced and sus- tained each other; but in neither state did any of the three influence the position of V in (1). I would infer from this that the fact that Boke and typapyéreve are both V did not suffice to make the 1 SEG, xm, 239 and DGE, 838, 23 ff. are the oldest examples (respectively eatly and mid fifth century B.c.) of the Argive formulae, which were still in use in the second century B.c. (e.g. DGE, 99, 2ff.).LOGICAL DETERMINANTS 57 Athenian or the Argive regard Sof ti} Boul}, «tA. and 8 Seiva typapudrreve as utterances of simjlar structure. The interaction of formulae can be followed in greater detail in the prescripts and subscripts of the Athenian Tribute-Lists. The factor common to all, which I omit, is &ml tijs dpyiis (with or without an ordinal numeral), to which § refers.’ ATL 2, 1 kat Aftév typouperreve? 351 Alt Aid[t}iuos typo udcreve] Since it was common practice to identify a board by naming its secretary—this identification, and not a desire to give interesting information about what Leon and Diotimus were doing, is the point of naming the secretary—# followed by a name makes typouucreve predictable and therefore C. This interpretation is supported by lists 4-8, where the secretary's demotic is added: 4,0 Affi....... Dis typappéteve Aat]Aipdarios 5,u Af .........0. typapy]écreu[e Ha}Acnev[s] 7,1 Ajé Mevér[ipo]s typalupérjeve Aou[trrpeds 8, 2f. AB Aiod[& éypopudccreve Monovi [Ses The pattern WC NV reminds us of dedication formulae, in which éviénxe is C. The situation in list 13 is very different: 13, rf. [Af XoA}miSeds MeArted[s eypoupdrreve Alo[p]uprvos "Iapreus heAAEVOTOULas Ev Z4[Tupos] AevKovods ouvel ypoupécreve] Here not only the secretary is named, but also the eponym of the board and the co-secretary. The fact that others besides the secretary are named automatically confers NV status on éypaypérreve because it becomes antithetical; hence XoAxiSels MeArteds typappécreve is NNN. 1am not arguing simply that ¢ypapypérreve must be inter- preted as NV because of the change in order; for lists 20-3 reveal the antithetical nature of £ypoppdrreve by chiasmus: 20, 1f. Ag Ofowidos ....... ] *Axop[ve}s eypoppdcreve Aea[pevofrfoptas & ....].[.e eee eee & Kepoy]ésv 21, 1f. Ag Tipatdvilkos & Kepoyéov ’Emycpos [typoumdrreve hedévorouilas Fv... . ulexos Xapiiguo XouTreraiov 22, 2ff. Adi] OAs.........- Jecto[......-655 typau[pérreve Asd)Agvor[aulas Ev Atjoviiaios 23, 34. Adi . .Juoxdpés Mup[p]ivfdoios typayyd{r]eve [Ae]AAEvo- Taptas By [Oi]AErenpos ['I}kofpret]s58 GREEK WORD ORDER The prescripts of lists 16, 17 and 19 are illegible, and list 18 is lost altogether, so that we do not know exactly when, between list 15 and list 20, the composer of the record chose to adopt chiastic order; but once adopted, the new formula was maintained until in list 25 (24 is missing) the character of the whole prescript was altered by naming all the members of the board. The serious problem lies in lists rx and 12: 11, 1 Adi Erpdp[Pryos Xo]AAelBes éy[pop]ucrreve 12, 1 Aét [ZJoplos éypafypér]eve *EAcvatvi(os) In list 1x the order which I have described as ‘antithetical’ is adopted; yet there is no antithesis, as no one else is named. List 12 reverts to the earlier practice, in which ypoypéreve is treated as C. The co-secretaty and the eponym are not named in the prescript, but separately, at the bottom of the list (12, 36), and they are stated in ‘antithetical’ order: Zétupos Aeukovoels youveypou[péreve Z]opoxAts KoaAd[vitey heAAévorouflos tv. The explanation of these phenomena is that formulae belonging to one series of documents influenced similar formulae belonging to other series. In the pre- scripts of decrees several individuals were named as performing different functions. The composer of the prescript of a tribute-list therefore had available to him three sets of models: (i) previous tribute-lists, (ii) other documents in which only one official was named, (iii) other documents, notably decrees, in which several officials were named and their different functions specified. Lists 11 and 12 are a battle-ground of models, and lists 13ff. represent the victory of the third set of models. The existence of formulae has an important bearing on the com- pilation and use of statistics. Suppose, for example, that we wanted statistical information on the relation between the imperatival infinitive and its object in the language of Attic documents. JG, 1°, 81 (421/0 B.c.), § opens with the words Tdv ‘Petrov Toy Trapé TS [Eloteos yepupSca1, and 94 (418/17 B.c.), 4f. with the words ipyocn ‘70 huppav 76 Ké8po . . . Kal prodScos TO Thevos Karr Ts cuvypaipdss, In the former case the object precedes the infinitive, and in the latter it follows. Now suppose that we add to these two examples the scores of decrees which record the public commendation of indivi- duals or states and begin, almost without exception, with the words trrenvéocn Tov Eelva. If we counted each of these commendatory ”LOGICAL DETERMINANTS 59 decrees as one example, we should conclude that it was much commoner in Attic decrees for the object to follow the imperatival infinitive than to precede it. In one sense, this is literally true; but it would not necessarily follow that the composer of an Attic decree, if required to say ‘strip the statue of its gold’, would be more likely to say TwepieAgoben Tol éydAperos Tév xpuody than ToU dy dAyerros tov xpuodv tepieAéoben. The Athenians often had occasion to commend people; they rarely had occasion to bridge the Rheitus or to fence the sanctuary of Codrus. Thetefore the expression of commendation became formulaic at an early date; and the more instances of the use of a formula we include in our statistics, the more we distort the picture which our statistics were intended to give us.’ ‘The influence of such formulae on each other may be seen.on a grand scale in decrees. An Attic decree usually makes the provision: ‘let the Secretary of the Council inscribe this decree on a stone stele’. This provision sometimes. begins with the words 1d 8 Whpicpa Tobe, sometimes with dvarypéyon St, sometimes with Tov 88 ypouporrta; by the end of the fourth century the form beginning with évorypéypon 5é emerges, after a varied career, with the status of a formula. Another common provision is ‘and invite the ambassa- dors (etc.) to dinner in the prytaneum’; and here too the formulation beginning with koAtoon 8¢ or Kal xoiAéoon has an almost unchallenged predominance.* To draw general syntactical.conclusions from these instances would be not only rash in principle but in conflict with the results of statistical inference from contemporary literature. If we could trace a formula back to its source we should find that the initial lead of one formulation over its alternatives was the product of what we, as students of the history of the language, would be compelled to call ‘chance’. Once the lead is established, the likeli- hood that this leading formulation will become a formula is greatly increased; so is the likelihood that it will help to determine the * Schwyzer, 01, p. 693, by going on from a distinction between ‘habituell’ and ‘okkasionell’ to cite many examples of the type Kexporrls érputaveve and add ‘Doch liest man auch Eyves B&pos Del.3 623 (Erythrae, II*)’, may give the impression that Kexpottis éputéveve exemplifies a general principle while Eyvw Bayos demands a special explanation. This impression would be fundamentally misleading, since the alternative formulations are syntactically indifferent and the predominance of either in a given community is a question of habit and tradition. 3 McDonald, pp. 153f.60 GREEK WORD ORDER ptedominant formulation of other utterances of similar structure and content. When these formulations conform to each other, their _.. very conformity strengthens the predominance of the formula with which the process began. Despite this, certain formulae have been used statistically to prove that one syntactical order is ‘normal’ and others ‘abnormal’. But to collect hundreds of instances of dedicatory inscriptions or artists’ signatures’ or proverbs,” and to argue syntactically from them, is illegitimate. The more instances we collect of identical formulation of a single type of utterance, the more conclusively do we demon- strate the existence of a formula and the less relevant does our information become to the establishment of general syntactical rules. Two utterances of type (a) tend to be formulated similarly simply because they are both (a). Every sustained utterance, colloquial or literary or administrative, is necessarily in some degree formulaic; it is hard to say anything which does not in some way resemble something which one has said or heard before. I set out below the various ways in which Homer, in the catalogue of ships, and Herodotus inform us of the name of the commander of a force or ship.3 I arrange the examples in an order designed to show (so far as is possible when one is compelled to operate in only two dimensions) the extent to which each example is derivative or original. Dots indicate words in apposition to the commander’s name, dashes words in apposition to the name of the force. (a) B727 GAAK MéBcov Kdopnoe .. . (6) B704 GAA opens KécuNGE ToSdpxns ... I, 103,3 tye St atrrods ... Mans ... VI, 92,2 ye St atrrols ... EvpuBderns ... IX, 17,2 frye 58 ovrrous ‘Appowuins ... VIL, 63 Fpxe 5€ opecov ’OTdoTs ... VI, 103, 1 Fyov 8 ogeas otparnyol SéKa VII, 222 do tparrtyyee St axtrrésv Anudoiaos ... IX, 28,6 éotporrijyee 8 cnirréiv “ApiotelBns ... IX, 96,2 totparrtyee St antrrod Tiypdévns ... * Wackernagel, Gesetz, pp. 430ff. ? Fischer, pp. 2ff., Barth, pp. 28f., asf. 3 T exclude such passages as Hdt. vu, 215, 1x, 66, 2, where the relative Téav means not kal toute bat TovTous dv,VIII, 131, 2 VII, 121, 3 Vil, 71 VII, 61, 2 Vil, 62, 2 Vil, 67, 1 (2) B 609 B 736 B 826 Bgy12 B 830 B 842 IV, 120, 3 Iv, 128, 2 VI, 211, 1 vil, 180 VIII, 47 VIII, 82, 1 vu, 181, 1 vu, 182 VII, 121, 3 VII, 194, 1 VI, 205, 2 B 713 B 586 B 576 @M y11 VII, 233, I (e) B68s5 B731 B 540 B 552 B 563 LOGICAL DETERMINANTS . 61 otpormyds 68 Kal vowapxos. fv Acutuxlins .. . otparmmyous 68 Tapelyeto EyepSopéved te Kal MeydéBugov" &pxovta 8 trapelyovto Macadyny ...* Kad &pxovta trapelyovto ’Otavéa ... trytudva Trapexopevor Meyérravov ... trysiidva Trapexdpevor ApropapSov ... tev Fpy’ ... "Ayorrivep éftxovra vetiv t&v Ap’ Evipirruaos ... t&v Apxe ... MdvBapos TEV Fpy’ "AaKdiAagos Kal "IdApevos ... T&v Apy’ “ABpnoTés te Kal “Auotos ... tév Fpy’ ‘Itrrd80ds Te TWAatds 7”... THis Fexe “ISaveupcos ‘THis Fpxe Zkcotraons? ‘iis Fipxe “Y¥Sdpuns ‘ijs Fipxe Tipn€ivos Tis Fipxe ... D&UAAos iis fipxe ... Tlavadrios ... tis Etpinpdpxes ’Accovidns tis Etpinpapyes Ddppos ... tis totportyeov Tprravratyuns Te Kad Pépyis aay totporrthyee ... LavSdoKns tétv Eotporijyee Acovmidins .. T&v Apy’ ... EvBexa vydv EGunros tev ot... fipxe ... Mevédcos éEtixovta vedy tev sxorrdv vydov fipye ... "Ayanéuvow ... ‘téSv 6 MeydPogos fipxe* tev & Acovndins totporiyes* Tév Of trevriiKovTa veddy Fv cpxds ’AXIAAEUS tev 00” tHhyeloOny ... ToBaAelpios HS Maydcov tév ot tyeudver’ ’EAgptiveop .. . Tésv o8” thyepdveve ... Meveobads tev ot” tHyepdveve . . . AtourSns Kal Léveros ... * Here there is explicit antithesis between the command of a force and some other aspect of it. T doubt the propriety of including these three passages in my list, since in all three cases (note the definite article in two of them) the purpose of the relative clause is not so much to inform us who the commander was as to remind us which force is being referred to.62 B Gor B627 B 740 B 837 B 622 Cf) B78 B 636 B758 B 657 B 698 B 650 B 620 B 870 B 678 (s) B 623 VI, III, I VI, 97 VII, 97 (A) vit, 66, 2 VII, 68 VU, 73 Vil, 77 VII, 80 ‘VII, 81 GREEK WORD ORDER tay até’ Hyeudveve ... Néorop tév at’ Hyeudveve Méyns .. . or tév a0’ fhyeudveve ... ToAutroftns ... té&v ave’... Fipy’ “Agios ... té&v 8... fipye ... Arcopns™ té&v BE Dirdoxrhtns fpxev ... Emre veddv tev piv *OSuccets fipxe ..- tév dv Tipddoos ... tryeudveve way iy Tren Gbos « «+ flyshvevs résv ot Tiporrectdctos . ~ frysubveve Tay v py dp! “IBouevets « + Hyeudveve Mnpidvns tév oubv & *Augleros kal OdAmOs Hynedoiny? tev bv &p’ ’Apgipayos Kal Néotns hynodoinv tev ott DelSitrirds te Kad “Avrigos hynodotny té&v BE teT&pToo fpxe ToAUEewes .. .* ‘rot pv BeEi00 Kkpeos thyteTo ... KaAAlpoxost Tiis BE GAANs otperrifis totporrtjyeov of Bo Tou 58 vauTiKoU Lotpartjyeov ol6e? _ Totrteoy & fipxov olBe? torrreov Bé fipxov olde? Tovtav ouvaygotépwy fipxe "ApTéyuns ... tovrroov Trévtov fipye BaSpns ... tote 58 Té&Sv vactwtiov fipxe MapSovrns ... toto dv Tol otpatot fipxov piv otro: .. .3 VII, 121, 3 Tonrns pév Bi) dotpartyeov MapBdvids re Ked vil, 83, 1 @ = vu, 79 G) vu, 82, 2 Maciorms* tév 8 puplov touTwvy —-- torpariyee piv “YBépuns3 tovrrev 5 Macionios ... fipye totparijyeov 8% touTwav —~— Mapbdvids te ... kod Tprravratypns3 (k) vil, 173, 2. dotpariyee St AoxeSoapovicov piv Evatvetost @ B8s8 B 645 Muodsy 58 Xpduis fipye Kad “Evvopos ... Koenrédv 8” "ISopevets ... tyyepdveve * Here there is explicit antithesis (with pév/5é), within a single complex sentence, between different forces. ? Followed by a list of names. 3 Here there is explicit antithesis between the command of a force and some other aspect of it.B 863 B 494 VIL, 725 2 B 517 B 856 (m) B 844 B-756 VII, 62, 2 VII, 66, 1 Vil, 67, 1 VIL, 7§, 2 VII, 64, 2 VII, 69, 2 VII, 74, 2 Vil, 79 Vil, 97 VII, 131, 3 B 638 B 851 B 527 B 819 B 816 VU, 69, 2 vil, 62, 1 vit, 67, 2 (x) B 862 B 537 B 748 B 671 B 653 B 867 LOGICAL DETERMINANTS 63 Mrjooi at Méc€Ans ‘re xal “Avtipos Aynod- Borerrdsv pév TinveAges Kal Atjitos Aipxov TlopAcryéveov yév vuv Kal Marrinvéiv Adytos ... fipxe? are Deokiiav ZxéBios kal "Eriarpogos fipxov anrrekp ‘AAizcoveov “OBtos Kal ’Emlotpopes fipxov avrrap Gpriixas fly’ *Akduos Kal Telpoos ... Mayvijreav 8° fipxe Tpd8oos Kigotoov 88 fipye "Avdens ... “Aploov 8 Fipxe Zicduvns ... Lopayytov BE fipye Depevdéerns .. Opnixcov 58 -—— Apye Bacodans ... Boxtploow && xed Zoxtoov Fipxe “Yor&orms . *ApaBloo 8 kad Al@idrranv ——— fipye ‘Avot. AuBév 8 Kod Muadiv fipye “Aptappévns . . Mapédv 58 Kal KéAyoov fipxe Sapevberns - Alyutrrioov 88 orrporrtjyee ’Axonnévns . "A@nvadeov 88 totpartyse ZdvOrriros .. . AltaaAdav 8 tyetro Odas ... TlapAaydvew 8’ tyetto MuAmpéveos Adciov Kijp Aoxpéiv 8 fyyepovevev ... Alas AapSaviav at’ fipyey ... Alvelag Tpwol uty tyendveve ... “Exroop ‘tév pv St Urrép Alyurrrou Alfiérreay kal "ApoBiav fipxe Apodunst ol 6. Mabou &pxovta piv trapelyovro Trypévny Tiéacrues 88 &pyovra trapelyovto *Aptativrny ... @édpxus a Oplryas fiye Kal ’Ackdvios . Alas 8 &x ZoAcuivos éyev SuoKaiBexcr vitos Touvels 8’ & Kugou tye Buca kal door vijas Nipevs oc ZUpnGev dye tpels vijas etoos TAntréAguos 8... & ‘Pddou évvéc vijas cyev Néortns at Kapéiv tytjoato —-— * Here there is explicit antithesis (with pév/Sé), within a single complex sentence, between different forces. * Here there is explicit antithesis between the command of a force and some other aspect of it.64 GREEK WORD ORDER (0) B631 ata 'OSucaels tye KepadAfivas ——— B 848 outap Tupatyuns &ye Matovas ——— B 840 ‘InerrdBoos 8 eye pic MeAasyav ——— B 876 Laprndev 8’ Apxev Auxioov kai PAadwos .. . A first glance over this list may suggest unlimited variety of formulation, but a second glance corrects the impression. Homer and Herodotus confine themselves to a proportion of the possible total of permutations.’ If we turn from information on the com- manders of troops to information on the names of persons and places, we find that Herodotus’s formulations of ‘whose name was ...’ in part correspond to his formulations of ‘commanded by’; otvona 8 of iv N~ tye 8 atrous 6 Siva and Tod otvopé tom V ~ tév Hpxe d Selva. Yet straightway we encounter a new formulation: 1, 179, 4 “Is otvoper arirrij, 1, 205, 1 Tépupis of fv obvope, 11, 29, 3 Toxopipds otvoper onrrij £oT1, VIIL, 32, 1 Tidopéa otvopte otrrij. This has no counterpart among the formulations of ‘whose commander was ...’; butif we seek its ancestry, we may find a cluein Hecataeus, fr. 282 by 8 ovrroton TréA15, Topikévn otvoyc. This brings us back finally to the ‘conflict of pattern and principle’ of which I spoke in ch. 1, (ii). In attempting to explain the word order in any given Greek utterance, we must ask not only ‘with what principles is it consistent?’ but also ‘what are its models and what is the history of its models?’ The part played by patterns and models offers an explanation of the process by which the syntactical * It should be remembered that the Homeric catalogue, unlike Hdt. vu, is simultaneously a catalogue of forces and a catalogue of heroes. * Barth, pp. 37ff., emphasises the importance of formal analogy and association. Frisk, p. 76, makes the point that an order originally determined logically may become ‘mechanisiert’ through familiarity and may eventually be employed in utterances to which the original determinants are entirely inapplicable. Elsewhere in his argument he perhaps underestimates the importance of this phenomenon. Thus (pp. 56ff.) he explains the order P Sin rdv fipxe 6 Selva: by saying that 7px is a‘ Rubrikwort’,and the exceptions X. An. 1, 7, 11 dv "Aptayépons Apxe, HG, 11, 1, 6, tv, 8, 10, VI, §, 11, by saying (p. 59) that for Xenophon (unlike Herodotus and Thucydides) fipye is not a ‘Rubrikwort’. What then is the explanation of X. An. rv, 8, 18 dv Fipxev Aloxlvns 6 *Axapvé, ibid. dv Apye KAscvoop 6 "Opxontvios, HG, 1, 2, 16? Tt might seem that f\pxe cannot both be and not be a ‘Rubrikwort’ for the same author in the same works. But perhaps it can, provided that we reverse the cause and effect. Herodotus chooses to make fipye a ‘Rubrikwort’ by putting it first; Xenophon sometimes chooses to do so, and sometimes not; but the word order cannot be invoked to explain the choice.