Anceps Devine, Andrew M;Stephens, Laurence Greek, Roman and Byzantine Studies; Summer 1975; 16, 2; ProQuest pg. 197 Anceps Andrew M. Devine and Laurence Stephens I NTRODUCTION. In current metrical theory anceps is generally assessed as a third autonomous element of the metrical structure distinct from longum and breve. In this paper we show that the ancient testimony appealed to in modern treatments does not sub­ stantiate this assessment, and that a third distinctive element is a priori improbable since it would find no parallel in the metrical systems of the world. In an explicit and formalized manner we apply certain techniques for the analysis of the organization of patterns (namely structural and transformational-generative analysis) which were developed primarily in linguistics, but which are applicable to patterns generally and hence to metrical patterns in particular, and we establish that (a) anceps is not a third metrical element, (b) it does not involve distinctive duration intermediate between longum and breve, (c) in the iambic trimeter metron and foot are not incompatible organizational concepts, and (d) no internally non-constrastive feet are basic higher level units of Greek metre. 2. ANCEPS. In Greek and Latin, as in various other languages (e.g. Finnish, Vedic, Arabic, Hausa, old English) there are two linguistically functional syllable quantities, and these playa part in the imple­ mentation of the metrical elements of the verse patterns utilized by these languages. These patterns are achieved by the obligatory exclusion of all but one of the metrical elements1 in given positions within a larger metrical unit. It is often the case, however, that at certain positions in the patterns2 and particularly at boundaries (e.g. in Greek, metron initially/finally when the realization of the basic foot contains two, but not when it contains three syllables; stichos finally), this exclusion rule does not hold, and the result is a free choice between two elements: as Marius Victorinus puts it, "nihil 1 In fact, whatever their linguistic realization, no metrical system yet scientifically examined in any language shows more than two distinct metrical elements, and we demonstrate below that Greek and Latin are no exceptions to this rule. 2 We are basically restricting our attention to podic metres in this paper. 197 198 ANCEPS metri interest utrum illa [final ancepsJ Zonga sit an brevis."3 In terms of the number of syllables or metrical elements anceps is a one-to-one variation, whereas other kinds of metrical variation involve either two-to-one (the mora-based resolution and contraction) or one-to-zero (catalexis, etc.). 3. FINAL ANCEPS AND ANCEPS PROPER. The above described phenomenon in classical verse has, since the time of the humanists,· been subsumed under the term anceps. Some modem metrists5 like to make a dis­ tinction between this free choice at the end of a stichos (final anceps, i.e. brevis in Zongo and the rare and disputed reverse thereof6) and anceps proper (e.g. metron initially/finally in iambics and trochaics respectively) : some such distinction does seem inherent in the different treatments accorded the phenomenon in these two positions by ancient metrists and grammarians, and we shall argue in favour of it below. Anceps proper may be treated as the result of one rule and brevis in Zongo as the result of a different one: the distinction thus drawn depends inter alia on the metrical elements that the rules change and, of course, on the environments in which they operate­ syZZaba brevis in elemento Zongo means that there is a rule optionally changing the long (final) element of the basic pattern into the short element. On the other hand, the structural effect of the two rules is identical, namely to produce alternation between elements; this congruence is reflected by the use of the blanket term anceps to refer to both rules. 4. ANcmNT TREATMENT OF ANCBPS. Ancient metrists prefer to use feet ratherthan smaller units as the basis of their analysis whenever possible, and therefore anceps proper in stichic metres is not specifically recog­ nized as distinct from resolution (by metrists as opposed to rhyth­ micians), but rather the different feet resulting from both processes are simply listed along with the iamb or trochee in parallel fashion. Hephaestion 15.16 (Consbruch): ".,. f3 ' <:" ", " , , .1 0 taJL LKOV O€X€TaL KaTa JLEV Tac 7T€pLCCaC xwpac TOVT€CTL 7TPW- , , "f3 'Q '<:' ~ <:, I \ 7"T}V. TpLrT)V. 7T€JL7TrT)V LaJL OV. TpLfJpaXVV. KaL C7TOVO€LOV. OaKTVI\OV. a This formulation does not, of course, preclude secondary conditioning or statistical tendencies in any such position. & L. Rossi, RFIC 91 (1963) 52. 5 e.g. P. Maas, Greek Metre (Oxford 1962) §34; A. M. Dale, The Lyric Metres of Greek Drama (Cambridge 1968) 26. 8 cf. Maas, op.cit. (supra n.5) §34 and note. ANDREW M. DEVINE AND LAURENCE STEPHENS 199 ) , '\ ~\ , " I ~, , avamXtCTOV, KaTa oE Tac apnovc TOVTEcn oEvTEpav, TETapT7Jv, ., "a "(:I , ) , EKT7JV taJLfJov Kat TptfJpaXVV Ka" ava7TatCTOV' For final anceps on the other hand a separate term exists at the level of the single element, namely fJ &Fu&4>opoc (cvAAaf11j) , syllaba indif ferens: the results of final anceps are described just as those of anceps proper in terms of foot variants, but with the important difference that here the &St&4>opoc is given as the motive for the variation. Hephaestion 16.2: "0 \ .,., '\ " , \ ~ \ , \" Q TE JLEV ovv aKaTal\TJKTov ECTLV, E7Tt T7JC TEI\EVTataC TOV taJLfJov ~ I I '" , ~ \ \ ,~ 'A.. OEXETa£ JLovov "I 7TVPpI,XtoV OLa T7Jv aOLa'(JOpov The more specific treatment of final anceps, as opposed to anceps proper, may find its explanation in the greater generality of the former. Hephaestion 14.15: n \ I , ~ 'A.. " • \ , \ \ .Q' " avTOC JLETpOV aota'(Jopoc ECTLV "I TEI\EVTaLa CVI\I\afJ7J, WCTE ~ , (} l' .' , Q ~ \ I ovvac a£ €tva£ aUT7Jv Kat, fJpaXELav Ka£ f.LaKpav since final anceps occurs in all sorts of metres and always in the same easily recognizable position (coda). However, some parallelism between anceps proper and final anceps does emerge from the phras­ ing of one scholiast, and Rossi's claim7 that &oufcpopoc can refer only to ''!'ultimo elemento del verso" requires qualification, for the term is used in a reasonably technical sense by the scholiast to Hephaestion ch. V, not only in the statement &ouj.cpopa yctp Tct TEATJ (116.20 Consbr.), but also in the statement 7Taca f.LETPWV &pX~ &Sufcpopoc (ib. 115.1). 5. ANCIENT TESTIMONY FOR INTERMEDIATE DURATION OF ANCEPS. Some modern metrists8 interpret anceps as representative of a metrical durational unit in the abstract scheme intermediate between the long and short elements, and in support of this view they have cited Aristoxenus §20 (Westphal): fin ~ \ - ~ -., ~ \ , \ " y_ \' , ~t!ptCTa£ oE TWV 7TOOWV EKaCTOC 'ITOL I\oy£[) nv" "I aAoytfi:' TOWUT'!l, " '" I \ I ,~ , 8 ' ,\ I JI I ~. "Inc DUO I\oywv yvWP£JLWV T'!l a£c 1JCEt ava JLECOV ECTaL. YEVO£TO 0 .. \' I ... ~ A.." \ A..(} , ~ , ,~ • \ av TO E£P1Jf.LEVOV woE KaTa'(JavEc E£ 1\1J'(J Et1JCav OUO 7TOOEC, 0 f.LEV )/ ,,, ...... '" \~I t I t~\'" "COV TO avw Tcp KaTW EXWV Kat DtC7Jf.LOV EKaTEpov, 0 OE TO JLEV KaTW ~, \~\"., I",' \A..(}' \ \ otC1Jf.LOV, TO oE avw 1JJLtCV, TptTOC DE TtC 1\1J'(J EtTl 7TOVC 7Tapa , \ \ Q' JI l' ~ 'A.. I JI \ t'\ TOVTOVC, T7JV JLEV fJacLV tc1JV av TOtC aJL'(JDTEpOtC EXWV, T7JV DE 7 Rossi, op.cit. (supra n.4) 61. 8 e.g. Dale, op.cit. (supra n.5) 7; W. J. W. Koster, Traite de metnque grecque l (Leyden 1953) 27; and A. Kolar, De re metrica poetarum Graecorum et Romanorum (Prague 1957) 44 and 91ff. 200 ANCEPS " , '0" ....." f \ ..... , apcw /LECOV /LEyE OC EXOVCaV 'TWV apCEWV. 0 yap 'TOtOV'TOC 7TOVC "\ '''I:' " ", " "' •• -'-\' ~I:' al\oyov /LEV ES Et 'TO avw 7TpOC 'TO Ka'TW· EC'TaL 0 7] al\oyLa /LE'Tu.s U "" \1 1 ~ '0' ~ " , '" , 1 OVO I\oywv YVWpLfUAJV 'T'[J ate 7]CEt, 'TOU 'TE LCOU Kat OL7Tl\actOV. KaAEi'Tat 8' OV'TOC xopEioc aAoyoc. Although, as usual in ancient discussions, no examples are cited, it is a reasonable assumption, going back to Boeckh and approved by Westphal,9 that the reference is to anceps phenomena. The only clear evidence from the tradition in favour of this assumption is from Aristides Quintilianus 37.24 (Winnington-Ingram) and Bacchius Senior, Elcaywy'lj 25 (Meibom); the latter cites a spondee as an example of a long thesis and an irrational (aAoyoc) arsis. The ancient testimony appears to be stating that in iambic and trochiac metres dactyls, anapaests and spondees change the durational ratio of avw to KCX-rW, i.e. --j except coda, with rare exceptions). This entails the assumption that the basic or most abstract representation of the hexameter is purely dactylic and that spondees are derivative result­ ing from the application of a rule. This has been the implicit assump­ tion of most metrists, but explicit formulation is really necessary for a proper analysis of hierarchical structure in less transparent metrical phenomena by the same technique. It is vital to recognize the difference between transformational rules and the rules of historical development: the former are theo­ retically separate from and make no claim about the latter. The assumption that the dactyl is basic in the dactylic hexameter does not imply that there was ever a period at which the hexameter had no spondees. Naturally transformational-generative rules and historical rules often coincide, but not always nor of necessity. 9. STRUCTURAL ANALYSIS: ANCEPS AND NEUTRALIZATION. It is generally realized that metrical free variation does not imply a separate ele­ ment in the case of resolution, for instance, where there is free varia­ tion between a long and two shorts. In fact, as will be immediately clear from the parallel situation illustrated below, sound structural analysis at any particular level will never lead to positing a separate, otherwise non-occurring, element to explain free variation between two elements: this would involve the fallacy of accounting for non­ contrastivity by contrastivity (in a single level analysis25 ). Take a table round which are placed stools and chairs in the follow­ ing pattern (depending on where you start counting): CHAIR - STOOL - CHAIR - STOOL/CHAIR etc. 123 4 An analysis of the above structure into occurring items leads to the assumption of two distinct units (the CHAIR and the STOOL), which we U M. S. Ruyperez, Emerita 23 (1955) 89. 25 In a multilevel analysis, of course, there would be no a priori objection to this pro­ cedure, provided there was some motivation for it. It will be clear from the discussion of transformational-generative analysis below that there is no motive in the case of anceps. ANDREW M. DEVINE AND LAURENCE STEPHENS 209 recognize from outside experience, and to the statement that in position 4 either may occur, i.e. the only thing demanded by the pat­ tern is some kind of available seat. It does NOT lead to the assump­ tion that in position 4 a third, separate, otherwise non-occurring unit is required, such as a lamp or a chair with a low back (intermediate between CHAIR and STOOL), for there is nothing in the existing arrange­ ment of chairs and stools that could suggest the occurrence of either a lamp or a chair with a low back. The only way to account for position 4 with one item is to say merely that a SEAT is required there; this instruction is then implemented by the choice of one of the two available types of seat, namely a stool or a chair. Therefore the pattern is: CHAIR - STOOL - CHAIR - SEAT etc. It is obvious that the difference between any SEAT and CHAIR is not comparable to the difference between STOOL and CHAIR: STOOL and CHAIR are both members of the class of SEAT, but CHAIR is not a member of the class of STOOL or vice versa. Free variation between two otherwise distinctive elements is not an uncommon phenomenon in linguistics. Its grammatical description depends on the overall framework adopted for phonostylistic and other variations (and in a generative phonology also on the degree to which the rules producing the free variation are morphologically or lexically constrained); nevertheless, the structural effect on an autonomous representation is one of neutralization, as pointed out already by Martinet in 1936.26 For example in Northern (British) English27 words like climb and clock are often pronounced tlimb and tlock, i.e. there is free variation between t and k before I at the begin­ ning of a word. Instead of either tor k (as in other phonetic contexts, e.g. time, cock), we have simply cany voiceless, non-labial stop', imple­ mented by t or k in free variation. Neutralization is a technical term denoting the non-distinctiveness in a subsystem of elements that are distinctive in the overall system.28 26 A. Martinet, Travaux du Cercle Linguistique de Prague 6 (1936) 55. 27 Example from W. Haas, Transactions of the Philological Society 1957 (1958) 150. 28 For a discussion see E. Buyssens, Proc. XI Int. Congo Ling. (Bologna, forthcoming), and Devine, Linguistics 109 (1973) 17-34. An intermediate duration for the longum in anceps position would simply constitute a contextually determined variant of the longum, which would still vary freely with the breve, with the structural consequences noted above. 210 ANCEPS In the case of anceps, free variation results in the neutralization of the contrast between longum and breve29 at least for final anceps and anceps in the pure alternating (i.e. iambic and trochaic) metres. Therefore the term anceps denotes an element unspecified as to quantity, just as any SEAT in our table example denoted an element unspecified as to the CHAIR-STOOL distinction, and non-labial voiceless stop a sound unspecified as to the dental-velar contrast. Such an element is known as an archi-element. The archi-element does not differ from the elements that implement it in the same way as they differ from each other. One thinks of Aristotle's ET€pa ... TO aUTO T£ oV'Ta (Met. 1018a), and if Aristotle's examples of this, i7T7TOC and av(}pw- 7T'OC (Top. I 103 a) , were lined up in the order HORSE-MAN-HORSE­ MAN/HORSE etc., one would hardly assume that the fourth element of this pattern was really an unavailable centaur. 10. GENERATIVE ANALYSIS: ANCEPS PROPER AND BASIC PATTERNS. It is surely not unnatural to suppose that a person assessing the arrange­ ment of stools and chairs around our dinner table would get the impression of a basically regular alternation between the two ele­ ments of the pattern, disturbed by an optional, subsidiary relaxation of the basic rule in every second CHAIR-STOOL sequence. He would thus be positing two levels or two patterns, a basic one with pure alternating structure, which is highly abstract, and a derivative one which accounts directly for the physical arrangement of chairs and stools as actually found, and which we have subjected to structural analysis in the preceding paragraph. To get from the abstract to the surface level, i.e. from the basic to the derivative pattern, we have the anceps rule, which optionally changes every second stool into a chair. We repeat that the basic pattern is an abstract one which is not necessarily ever implemented directly without the anceps rule, and does not necessarily represent an earlier form of the pattern in the sense of its historical evolution or of its genesis: i.e. the assumption of a basic theoretical pattern CHAIR - STOOL CHAIR - STOOL etc. (trochaic type) STOOL - CHAIR - STOOL - CHAIR etc. (iambic type) 19 Of course, this sort of neutralization is much less basic than the harmonic type of redundancy entailed by the repetition of the same foot throughout the stichos. ANDREW M. DEVINE AND LAURENCE STEPHENS 211 does not imply that this arrangement is now or ever was actually used around any table. Whereas we have seen in the previous section that structural analysis at the surface level is based not on intuitive preference but on principles of logic (which exclude the possibility of a third surface element), at the abstract level on the other hand no completely rigorous means of deciding between alternative proposals is always available (although of course there are sound criteria as we shall see below). Therefore, since the traditional anceps cannot be a surface element, it might seem feasible to defend it as an abstract element; but there are a number of objections to this, too. In the first place it is worthwhile to postulate an abstract pattern different from the surface pattern in order to capture the posited basic, pure alternation, but an abstract pattern with a separate third element does not seem to have any immediately obvious advantages, since free variation at the sur­ face is not necessarily best accounted for by a separate abstract element: such a solution would not be preferred, or even occur to most people, in the parallel case of the CHAIR/STOOL variation above. The real motive for the traditional third element hypothesis lies in the supposed evidence of performance. But, even if it were absolutely certain that in performance long syllables were or could be shortened to a value between long and short in anceps positions, this would not lead to the assumption of an abstract compromise element. Since, as noted above, there is evidence ex silentio that in performance short syllables were not lengthened to intermediate duration in anceps positions, it is clear that the posited pattern-motivated shortening of long syllables would be due to an attempt to realize directly the underlying SHORT element posited by us and not the underlying compromise element posited by traditional theory. There would be an interesting parallel for this in the performance of the Serbo-Croat epic deseterac, where, to maintain isochrony, the ninth syllable is phonetically lengthened in compensation for the preceding quanti­ tatively pyrrhic foot. We therefore suggest that the following is an intuitively satisfactory, although of course not empirically confirmable, transformational­ generative analysis of an initial portion of a Greek iambic stichos: the implication of this derivation is that it represents for any audience the psychological organization and assessment of the structure inherent in the metrical pattern in question. 212 A. ABSTRACT PATTERN U-U-etc. ANCEPS 1. ANCEPS TRANSFORMATION RULE U -+ - (odd-numbered feet, optional)30 [B. SURFACE PATTERN X - U - (where X denotes the archi-element produced by the above rule) 2. ARcm-ELEMENT REALIZATION RULE X-+{~]31 C. ACTUALLY OCCURRING INSTANCES OF THE PATTERN --U- U-U- The number of different permissible instances of the various metrical patterns in Greek is relatively constrained (as opposed, for instance, to the number of different verses that could be written according to those patterns, which is, of course, infinite). One possible metrical grammar therefore would be a simple list of all permissible sunace instances of each pattern (iambic trimeter, dactylic hexa­ meter, etc.). This, however, would be rather cumbersome, since there are about fifteen hundred different varieties of the comic trimeter.32 But there are more important objections to such a gram­ mar than the fact that it is an inefficient and uneconomical way of providing the necessary information. In the first place it suggests that a poet or audience judges the metricality of a line by checking through a list of hundreds or thousands of possible structures. And secondly it quite fails to relate the superficially distinct but basically similar and to separate the basically distinct but superficially similar: this is imposed on a metrical grammar as a primary task by the phenome­ non of responsion. A transformational-generative account of metrical patterns, on the other hand, meets all these objections, and directly incorporates the fundamental principle of responsion: all sunace patterns (correctly) derived from the same basic pattern respond. By the above metrical analysiS, furthermore, the old conflict 30 For the nature and function of optional rules see Devine and Stephens, New Approaches to Greek Metre, to appear 1976. 31 For the sake of Simplicity we have omitted other rules such as resolution. 32 To be precise 1510: there are 2250 mathematical combinations of the permissible foot variants, 740 of which are ruled out by the co-occurrence restriction 'no contiguous re­ solved elements'. ANDREW M. DEVINE AND LAURENCE STEPHENS 213 between the metron and the foot in pure alternating metres is resolved: for the metron only comes into existence as a structural unit at the level of abstraction of the anceps rule, prior to which there is only the foot, as in the dactylic hexameter. It is thus possible to assume that the iamb is basically a self-repeating foot33 just as is the dactyl, and that both of them are modified by subsidiary rules (anceps and contraction respectively) to produce their surface patterns: the difference is that while contraction applies to every dactylic foot out­ side of the coda, the anceps rule applies only to every other foot in pure alternating metres, whence the generation of a unit larger than the· foot (the metron) on the basis of internal responsion. It is well to note that in the above scheme Band 2 are logically demonstrable from (the empirically observable) C; but A and 1 are hypotheses which we find intuitively more satisfactory and valuable for an economical account of the diverse surface forms than other possibilities (such as --+u/ odd numbered feet, or some even more fantastic construct) on the grounds that they (A, 1) reveal internal responsion. In general, criteria for the assessment of conflicting abstract representations and associated rules refer to factors such as responsion, statistics, 'Qckham's razor' (i.e. no absolute neutralization, or in other words no abstract forms which never appear at the surface), naturalness of representations, naturalness of rules, etc. It is interest­ ing to note that Horace (Ars Poetica 253ff) seems to come out explicitly in favour of an analysis comparable to that just given, except that he confuses abstraction with historical priority, an erroneous procedure in which Horace is joined on occasion by certain modern linguists. He describes the iambic trimeter as consisting of nothing but iambs (primus ad extremum similis sibi) and explains that spondees were apmitted by a later innovation (non ita pridem34 ••• spondeos stabilis in iura paterna recepit). The obvious historical falsity of this claim for an original trimeter with no anceps, whether Greek or Latin, has disturbed scholars for centuries. What Horace has done, as just re­ marked, is to confuse transformational-generative analysis with historical development by assuming that the basic abstract pattern 33 A natural impression; cf. e.g. Jusatz (op.cit. [supra n.14] 268): "ut arsis, quam brevem esse rhythmus legitimus postulet, longam excipiat." 84 The various difficulties presented by "non ita pridem" in no way obscure the obvious fact that Horace is speaking of priority in the historical, genetic sense, which is all that our point requires. See C. O. Brink, Horace on Poetry (Cambridge 1971) 298. ad loco for bibliog­ raphy. 214 ANCEPS of the iambic trimeter (which he deserves credit for discerning explicitly) also represents the earliest stage in its evolution. Orelli already perceived the source of the confusion when he paraphrased Horace as follows: "'Quod ad 8€wplav et legem huius versus attinet' ait