Stephen Halliwell, et al. (edd.), Aristotle, Poetics. Longinus, On the Sublime. Demetrius, On Style. Cambridge, MA: Harvard University Press; Loeb Classical Library, 1996. Pp. 533. $16.95. ISBN 0-674-99563-5.
Reviewed by Heinz-Günther Nesselrath, Institut für Klassische Philologie, University of Bern, Switzerland.
This is a 're-make' of a Loeb volume first published in 1927; the critical editions of the Greek texts as well as the translations contained in that volume look now rather dated, and so it was a very good idea to have all of its contents re-examined and re-established by scholars of today. The result -- to put this quite clearly already at the beginning -- is a very useful book for everyone interested in three of the most important texts of ancient literary criticism. All three editors/translators have long been experts on their respective authors: Halliwell (henceforth H.) not only has published a monograph and several articles on the Poetics,1 he has also once before already brought out a translation of it. Donald Russell (from now on R.) has edited On the Sublime already twice: once in his still invaluable commentary of 1964 and for the OCT series four years later2; he brought out a translation in 19723 and published critical reconsiderations of his earlier work in 1981.4 Doreen Innes (hereafter I.) has been working on Demetrius' On Style since her doctoral dissertation (which, as far as I know, unfortunately still remains unpublished); she has translated On Style (with some omissions) for the same volume in which R.'s translations of On the Sublime appeared (see n. 3), and she, too, has published several studies on Demetrius and related matter of Greek literary criticism.5 So the reader of this Loeb volume may expect solid, reliable work from the outset; and he is not disappointed.
All three scholars preface their authors with a concise and informative introduction, providing a general outline of the following work, its characteristics, remarks on the establishing of text and translation and a useful selective (and very up-to-date) bibliography. R. and I. also give a detailed "analysis" or "synopsis" (respectively) of the contents of their text, while H. restricts himself to elucidating Aristotle's original and basic concept of the Poetics and how it came to be centered mainly on tragedy; he adds shorts discussions of some of the main notions to be found in the Poetics (elevated action, transformation, pity and fear, Hamartia, and -- inevitably -- catharsis). Generally, these discussions are very helpful, only the last one (on catharsis) remains somewhat vague (H. defines catharsis as a "concept which is interconnected with various components in Aristotle's theory of tragedy, and which in some sense completes his account of the genre by framing the experience of it as psychologically rewarding and ethically beneficial", p. 19); but then you really can't expect a definitive nailing-down of a concept which has tantalized innumerable scholars for many generations in an introduction of this kind.
The other two editors/translators have some particular problems of their own to deal with in their respective introductions, and they do it well: R. gives a carefully balanced summary of the questions of date and authorship of On the Sublime (p.145-8); he settles for a date "in the first century A.D. It is harder to be more precise" (p. 147), and he seems inclined to at least consider the possibility that the author might actually have born the name Dionysius Longinus,6 as he is called in the one basic manuscript on which the text depends (p. 148); this may explain, too, why R. never refers to his author in the usual round-about way as "Pseudo-Longinus" or "the author of On the Sublime", but simply calls him Longinus (in 1964 he had used the intentionally "non-committal" capital letter L). In the case of the treatise On Style, the author (whose transmitted name Demetrius, though, has itself never been called into question) proves similarly elusive, and I., in her discussion of date and authorship (p. 312-321), has to settle for a kind of split (and therefore not wholly satisfactory) answer (but again this may be the best that is attainable): "the contents at least do not preclude and may best reflect the second century B.C." (p. 313); "there may ... be a few points of language to suggest a date of composition as late as the early first century B.C." (p. 321).
Now for the texts themselves, all three of them accompanied by a selective apparatus criticus. H. states that his text "has been broadly based on the edition by Rudolf Kassel", but that he has sometimes preferred different readings and that he has tried to print "as 'clean' a text as possible, minimising such things as editorial brackets" (p. 21; indeed, I haven't found any in his text). It remains to be seen, whether such a policy is advisable in all instances; even after more than two thousand years of classical philology, there is no such thing as a really "clean" Greek text, and every reader might be better off, if he is told about places where a solution has not yet been found or at least remains very doubtful (in the same volume, I. has not shunned editorial cruces, and she is right in doing so). As to H.'s deviations from Kassel's text, there may indeed be some places where an alternative may reasonably be considered7; but in the majority of cases where H. has printed something different from Kassel's OCT, his new text -- to me at least -- does not seem to be an improvement. Examples:
In 1447a28-b2, H. edits H( DE\ MO/NON TOI=S LO/GOIS YILOI=S H)\ [instead of Lobel's/Kassel's <KAI\> H(] TOI=S ME/TROIS KAI\ TOU/TOIS EI)/TE MIGNU=SA MET' A)LLH/LWN EI)/Q' E(NI/ TINI GE/NEI XRWME/NH TW=N ME/TRWN A)NW/NUMOS TUGXA/NEI OU)=SA [instead of Kassel's A)NW/NUMOI TUGXA/NOUSI] ME/XRI TOU= NU=N; by this he makes Aristotle speak of one single art, regardless whether in prose or in metres, while subsequently Aristotle quite clearly gives separate instances of the two (1447b9-13). In 1448a23f. H. should have kept Kassel's cruces around TOU\S MIMOUME/NOUS; the words seem like odd intruders, and H. doesn't really translate them. Regarding 1450a12-14, H. at least acknowledges in a footnote that "text and sense are here greatly disputed", but prints a text (without Kassel's obeloi, and introducing two conjectural changes), in which the second sentence contradicts the first.8 As to 1450a17-20 (KAI\ EU)DAIMONI/A ... H)\ TOU)NANTI/ON), Halliwell should at least have signalled to the reader that these lines were excluded by Kassel; though H. defends them in his 1987 translation as "a clear link between Ar.'s view of dramatic action and his general ethical philosophy" (p. 67), they are not without problems in the textual tradition and seem -- in this context -- rather superfluous. In 1450b8-10 (E)/STIN DE\ H)=QOS ME\N TO\ TOIOU=TON O(\ DHLOI= TH\N PROAI/RESIN, O(POI/A TIS [E)N OI(=S OU)K E)/STI DH=LON H)\ PROAIREI=TAI H)\ FEU/GEI]), H. keeps O(POI=A/ (transmitted by A and B) as well as E)N OI(=S -- FEU/GEI deleted by Immanuel Bekker and bracketed by Kassel; this leads to an awkward repetition in the very next clause. H. tries to overcome this by translating E)N OI(=S OU)K E)/STI DH=LON with "when otherwise unclear" and adds in a footnote "sc. from the action"; but as we are dealing here with tragedy, where everything is represented by speech, this "otherwise" loses its sense. In 1450a12, not much is gained by the "unbracketing" of TW=N ME\N LO/GWN, nor in 1451b32 by re-integrating KAI\ DUNATA\ GENE/SQAI; as Aristotle has already stated a few lines before that everything that has happened must be something possible (1451b17f.), he would be contradicting himself if he now stated that some things that happened are possible; and as he is now concerned with probability (as something to be depicted by poetry) and not possibility, KAI\ DUNATA\ GENE/SQAI has to go.9 In 1452a35 simple excision of the words E)STI\N W(/SPER EI)/RHTAI SUMBAI/NEI still seems better than their labored resurrection with the help of Spengel's conjecture. In 1454a22, H. should at least have mentioned Kassel's A)NDREI/AN -- which in his 1987 translation he still followed -- instead of A)NDREI=ON, which introduces a noticeable clumsiness into the sentence (as his new translation shows). 1454b14f.: if one really is to make some sense out of PARA/DEIGMA SKLHRO/THTOS OI(=ON TO\N *A)XILLE/A A)GAQO\N KAI\ *O(/MHROS (bracketed by Kassel), Lobel's transposition of PARA/DEIGMA SKLHRO/THTOS after KAI\ (adopted by H.) is not enough; one might consider something like KAI/PER PARA/DEIGMA SKLHRO/THTOS O)/NTA, but one would be hard put to explain how then the transmitted text came to look like it does. 1459b5-7 (PLE/ON O)KTW/ -- *TRW|A/DES): Not one word that this enumeration of tragic titles is quite beside the point of Aristotle's argument and therefore probably spurious (it was so regarded by H. himself in his 1987 translation). In 1460a34f. the words A)\N DE\ QH=| KAI\ FAI/NHTAI EU)LOGWTE/RWS E)NDE/XESQAI KAI\ A)/TOPON seem to be a well-nigh hopeless case; in his translation, H. tries valiantly to make sense out of them, but he has to add some expressions which are simply not there ("If a poet posits an irrationality, and a more rational alternative is apparent, this is an absurdity"), and the whole doesn't fit too well into its context. Moreover, in his earlier translation, H. tried just as valiantly and got a completely different translation out of this jumble ("But even absurdity can sometimes be handled more or less reasonably"10); all of which very much justifies Kassel's obeloi. In 1461b9, H. has dropped Kassel's cruces around EI)KO/S E)STIN, but the resulting expression is totally out of character with Aristotle's normally crisp diction. In 1461b18, H. assumes AU)TO\N (obelized by Kassel) to be sound, but has to translate something ("so that the poet himself contradicts) which is not really in the text; he should at least have mentioned Kassel's conjectural replacement of AU)TO\N by E)NANTI/ON. In 1462a16, Spengel's deletion of KAI\ TA\S O)/YEIS (adopted by Kassel) still makes sense; just in the preceding sentence, Aristotle has stated that tragedy may reach its full impact just by reading, without any movements by actors, and in 1450b15-20 (to which H. himself draws attention in a note) he has even declared that O)/YIS is not really a poetical part of the art of tragedy -- and should he now argue the superiority of tragedy with regard to another poetical art (epics) by pointing to O)/YEIS? Moreover, if KAI\ TA\S O)/YEIS is deleted, changing the subsequent DI' H(=S into DI' A(/S (so Vahlen, followed by H.) becomes unnecessary.
As for the text of On the Sublime, R. has apparently rethought his former text very extensively; all in all, I noted about 80 differences from his OCT (which is virtually identical to his 1964 text). Only once R. has put in obeloi (in 10,2 at the end of the Sappho quotation), while in his former editions there are many more; and, as in H.'s case, it might have been advisable not to let the text appear more sound than it really is. In many cases R. has now rejected emendations formerly adopted and returned to the manuscript tradition; in others he has done the reverse. Many of his changes look like improvements, but not all of them.
Improvements by returning to the tradition: 5 (l. l. 5) E)PI/FORON instead of von Arnim's E)PEI\ FORO/N; 8,2 (l. 2) KAI/ (deleted by Pearce) before E)/DOCEN; 11,2 (l. 7) A)/RA NH\ *DI/A H)\ instead of A)/R' H)\ NH\ *DI/A (R.'s own conjecture!); 16,2 (l. 8) no more KATA/ (an early conjecture) before TW=N A)RISTE/WN; 17,1 (l. 8 and 9) no more obeloi around KAI\ TAU=Q' ("especially"), no more PA/NTAS TOU\S before E)N U(PEROXAI=S; 20,1 (l. 6) no more E)/XONTA (R.'s own conjecture!) before TA\ A)SU/NDETA; 22,1 (l. 3) no more lacuna (Wilamowitz) between OI(ONEI\ and KARAKTH\R; 22,2 (l. 5) OU)=N instead of Spengel's GA/R; 26,1 (l. 2) KAI/ (deleted by R. himself) before POLLA/KIS; 29,1 (l. 1) TO/ (deleted by Weiske) before PRA=GMA (the sentence becomes a bit more rugged in this way, but this may well be the original tone); 34,2 (l. 12) no more E)/XWN (added by Selb) after KWMIKO/N; 42 (l. 6) no more obeloi around E)P' EU)QU/ (so that a characteristically terse expression of our author results);
-- by introducing a conjecture: 2,1 (l. 2)11 Upton's PA/QOUS instead of BA/QOUS; 4,5 (l.4) Reiske's KATA\ TO/ instead of KAI\ TO/; 7,1 (l. 5) Reiske's OU)D' (not cited in R.'s former editions) instead of OU)K; 10,3 (l. 5) Weiske's deletion (not cited in R.'s earlier editions) of H)\ GA\R FOBEI=TAI H)\ PAR' O)LI/GON TE/QNHKEN, a neat solution; 10,7 (l. 8) EI)S (inserted by Roberts) before MEGE/QH; 15,4 (l. 4) KAI/WN (Richards) instead of KA/TW; 31,1 (l. 3-5) Vahlen's KAINO\N E)PAINETO/N instead of KAI\ TO\N E)PH/NETON (after that, R. puts DIA\ TO\ A)NA/LOGON -- DOKEI= between parenthetical dashes, a quite attractive solution); 34,2 (l. 16) Buecheler's TI after DIECODEU=SAI (instead of E)/TI); 34,4 (l. 8) Richards' KAI/RION instead of KU/RION; 40,2 (l. 6) von Arnim's DEO/NTWS (not cited in the former editions) instead of D' O(/MWS (compare 42,1 l. 4); 43,4 (l. 8) AU)TA\ KAI\ R(HTW=S (Richards) instead of AU)TA/RKH OU(/TWS; 44,1 (l. 2) KAI/ (an old conjecture) between E)PIPROSQH=NAI (Manutius for -QH=NAI) and DIASAFH=SAI; 44,7 (l. 8) EU)QU/S (Matthews) instead of EI)S A(\S;
-- by exchanging conjectures: 7,2 (l. 2f.) Manutius' PARA/STHMA instead of A)NA/STHMA (apogr., Ruhnken), supported by 9,1 (TA\S YUXA\S ... W(/SPER E)GKU/MONAS A)EI\ POIEI=N GENNAI/OU PARASTH/MATOS);
-- by a new combination of tradition and conjecture: In 9,14 (l. 4) R. restores E)K *KI/RKHS (instead of Faber's E)N K.) and adopts Valckenaer's SUOMORFOUME/NOUS (not noted, however, in the apparatus) instead of SOUFORBOUME/NOUS; in 27,2 (l. 8) he replaces Roberts' A)PO/LHSQE with Cobet's A)POLE/ESQE (without, however, informing the reader of this conjecture) and then restores the transmitted TRW/SETE (instead of Roberts' TRW/SHTE); in 43,1 (l. 8-9) he restores DRASSOME/NOUS (conceivable), replaces A)XA/RISTON by Herodotus' A)/XARI (much less conceivable after keeping the un-Herodotean DRASSOME/NOUS) and exchanges <KAI\> I)DIWTIKO/N (Roberts) for )DIWTIKO\N <O)/N> (Wilamowitz).
The following changes seem no improvement: 4,2 (l. 10) no more E)N (Cobet) after ME/N (but compare the following line); 4,4 (l. 15) no more lacuna between O)FQALMOI=S and I)TAMO\N (in a footnote, R. still acknowledges the transition to be "very abrupt"); 4,7 (l. 5) OI( (deleted by Wilamowitz) before BA/RBAROI (but the speakers are not all the barbarians, but simply some barbarians, as in R.'s own translation of 1972, and the following KAI\ E)N ME/QH| links up rather awkwardly with OI( BA/RBAROI); 9,12 (l. 6) PROEGNWSME/NOUS instead of Reiske's PROEGNWSME/NOIS (rightly praised by R. in the 1964 commentary); 13,4 (l. 2) H)QW=N H( (as in the 1964 edition, while in the OCT R. had obelized the transmitted H)QW=N H)/, which still seems the best solution, as R.'s translation doesn't really sound satisfying, and he himself considers an alternative in a note); 14,2 (l. 8) PEPAI=XQAI instead of R.'s own PEPLA/SQAI12; 16,3 (l. 17) O(/RKWN (deleted by Kayser) before PI/STIS; 16,4 no more obelos before KAI\ O)NO/MATA (but the translation -- rather different from the one of 1972 -- lets the uncertainty of the text still come through). In 31,1 (l. 1-2), as the text resumes after a loss of four leaves,13 it is indeed impossible to tell how the first sentence has to be divided: TO\ D' *A)NAKRE/ONTOS may be sound, but no one can say whether the subsequent OU)KE/TI belongs to it or to the following quotation, so at least OU)KE/TI should better remain obelized. 32,5 (l. 30): Can A)RAIOU= (un-obelized as in the 1964 text) ... AU)LW=NOS really mean "a conduit full of passages"?14 32,8 (l. 1-2) no more obeloi around O(/MWS AU)TO\ KAI\, but the words still seem redundant15; 32,8 (l. 6) no more deletion of the second TW=| PANTI/ within a few lines; 35,3 (l. 1-2) TH=| QEWRI/A| KAI\ DIANOI/A| TH=S A)NQRWPI/NHS E)PIBOLH=S instead of Ruhnken's TH=| QEWRI/AS KAI\ DIANOI/AS TH=S ANQRWPI/NHS E)PIBOLH=|, but the restoration does not advance the intelligibility of the sentence (in R.'s translation, "the speculative intelligence of human thought", the term E)PIBOLH/ is not rendered); 38,4 (l. 7) O(MOI/WS instead of Schurzfleisch's O(/MWS (which still gives a better sense). In 39,4 (l. 11), deleting TO/ TE before E)PEI/TOIGE probably does the text much less justice than R.'s former assumption of a lacuna. 40,1 (l. 2-5): If we accept R.'s placement of a colon (rejected in his 1964 commentary) before OU(/TWS in l. 5 and the Loeb translation of the whole period, we ought to put an additional comma after KAQA/PER TA\ SW/MATA.16 In 44,6 (l. 12) Spengel's GA/R before NO/SHMA is not necessary, if you put a comma (instead of a colon) before FILARGURI/A.
Apart from all this carping, however, this is a text where the editor's meticulous rethinking of every phrase and sentence gives one a lot to ponder; it is a pity, though, that this text is not free from misprints, some of them rather serious,17 while the other two are better in this respect, though not faultless.18
Last not least, the text of On Style. It may be worthwhile to compare it with the French edition recently brought out by P. Chiron,19 because the two differ quite significantly in their understanding of the manuscript tradition: While Chiron believes that two manuscripts, P (Parisinus gr. 1741, 10th cent.) and M (Marcianus gr. 508, later 14th cent.), are basically independent of each other (in Chiron's eyes, M even presents a fuller and more correct text), I. is convinced that "M is ultimately an idiosyncratic descendant of P" (p. 334) and therefore only to be regarded as a source of (sometimes) interesting conjectures. With these differences of opinion, it is to be expected that text and apparatus of both these editions vary considerably. The proof of the pudding, however, is in the eating; if one compares both editions, one may well come to the conclusion, that I.'s position is the right one: In no instance does her text appear inferior to that of Chiron,20 and in quite a number of cases it even seems rather superior -- not only because she bases her text on P, but also because she is generally more willing to consider conjectural emendation as a means for improving the text. Examples:
In 9 (l. 4), TE in TO/ TE "DIONU/SIOS E)N KORI/NQW|", which introduces an illogicality after TO\ PROEIRHME/NON (only DIONU/SIOS E)N KORI/NQW| is in fact PROEIRHME/NON, but TE would couple it with two further examples), is rightly deleted by I. (following Hahne); Chiron leaves it untouched, but translates, as if it wasn't there. In 42 (l. 2), I. posits a lacuna before OU)DE\ EU)/RUQMOS, arguing that otherwise heroic verse would here be restricted to spondees; Chiron sees no lacuna and thinks that for Demetrius 'heroic metre' or 'heroic rhythm' could really have meant only spondees, which might be very hard to prove. In 50 (l. 6f.), Radermacher's deletion (adopted by I.) of OI(=ON KATAPEPTWKE/NAI A)PO\ I)SXUROTE/ROU E)PI\ A)SQENE/S rids the text of a laborious and unnecessary explanation. In 52 (l. 3), E)PANIO/NTI ... <E)/OIKEN> (adopting an old conjecture) gives the text a typically Demetrian turn of phrase, while Aujac's E)PANIW/N TI (adopted by Chiron) introduces a meaningless TI and leaves E)PANIW/N (after KAI/) somehow dangling in mid-air. 57 (l. 1): P's ungrammatical PAQHTIKOI=S is easily changed into PAQHTIKW=S, and this is well attested in Gregory of Corinth's quotation of this Demetrius passage (so I.); M's E)PI\ PAQHTIKOI=S (adopted by Chiron) looks like an effort to correct P's reading. In 58 (l. 3) Nauck's E)PILE/GOUSIN (adopted by I.) makes good sense out of P's E)/POS LE/GOUSIN, while Chiron's PRO\S OU)DE\N E)/POS LE/GOUSIN (this is also the text of Rhys Roberts' older Loeb) is clumsy. In 93 (l. 7), I. well-advisedly adopts Spengel's DEI= POLLA/ (for the transmitted DIPLA=), which Chiron, again, does not even cite. In 95 (l. 6) Rutherford's insertion of O( before O)NOMATOURGW=N makes good sense; Chiron has no such insertion, but translates as if the article was there ("le créateur de mots"). At the end of the chapter, I.'s positing a lacuna shows her close attention to the development of Demetrius' argument. In 96 (l. 4) I.'s adoption of Rutherford's METACU\ *E(LLHNI/ZWN TOI=S O)NO/MASI makes excellent sense, while Chiron clings to the manuscript reading (of both P and M) METACU\ *E(LLHNIKOI=S O)NO/MASI and translates something which isn't in it ("tout en usant de mots grecs"). At the beginning of 128, Kassel's insertion of O(/S (adopted by I.) before XARIENTISMO/S fully vindicates the original reading of P (and M), whereas Chiron has to take refuge in an altered (and rather clumsy) version of the sentence standing in the margins of both manuscripts, obviously a conjecture. In 142-143, I.'s adoption of the conjectures of Finckh (KATAULEI=. H)\ instead auf KATAUDEI/H) and of Wilamowitz (change of speaker after MELANOPTERU/GWN and reading PURROPTERU/GWN instead of PRO\ PTERU/GWN) makes her text markedly superior to that of Chiron in this passage. In 151 (l. 9), Victorius' MIMIKW/TERA (adopted by I.) certainly suits its accompanying KAI\ AI)SXRA/ much better than the transmitted MIMHTIKW/TERA (kept by Chiron). In 164 (l. 6f.), the words W(/SPER E)/XEI -- GE/GONA, presenting a quotation already dealt with in 144, where neologisms were the subject, are rightly cut out by I. (following Hahne), as this passage is concerned with O)NO/MATA KOINO/TERA. Chiron sees the contradiction (cp. his note), but does not take appropriate action. In 172, I.'s insertions (TO/ in l. 1 before ME/N, following von Arnim, and A)POKALOU=NTES TO/N -- with A)POK. coming from Radermacher and TO/N from I. herself -- before MAKRO/N) restore grammar and sense; Chiron tries to make do (more or less) with the transmission. In 180 (l. 6), the transmitted AU)TW=N (kept by Chiron, but not translated) after H(MW=N is superfluous, Kroll's AU)TA/ (adopted by I.) very apt. In 195 (l. 3), von Arnim's deletion of TW=| O)/RNIQI (adopted by I.) removes an awkward gloss, which is kept by Chiron, but not translated. In 204 (l. 3 and 4), by deleting H(RWIKW=N (Spengel) and O)/N (Radermacher, Roberts) I. establishes a clear and intelligible sentence, whereas Chiron, keeping H(RWIKW=N and reading M's O(/ (instead of O)/N), is reduced to some quite tortuous logic. In 232 (l. 7), a vivid and pointed expression is gained by adopting Cobet's A)PO\ MHXANH=S (as I. does), while by keeping the transmitted MHXANH=S one would have to understand DIA\ MHXANH=S, and what would that mean? Chiron translates LALEI=N DIA\ MHXANH=S with "s'exprimer trop ingénieusement, avec trop d'art et d'habileté"; but there seem to be no parallels for this expression, nor does it go well with the preceding DI' E)PISTOLH=S. In 250 (l. 3), M has GRA/MMATA after E)DI/DASKES in the Demosthenes quotation, but this supplement (although E)DI/DASKES GRA/MMATA is the original text) is quite unnecessary for what Demetrius wants to demonstrate, and in other instances he doesn't keep to the entire original wording either; so I. is probably right in regarding GRA/MMATA as an addition by M and not (as Chiron does) as Demetrius' own text. A similar case is 279 (l. 5), where Chiron, following M, adds KAI\ PARESPO/NDEI (as in the original Demosthenes text), but I., following P., leaves it out. In 271 (l. 3), TOU=T' E)/STI DEINO/THTA is another gloss rightly deleted by I. (following Radermacher) and wrongly kept by Chiron. In 280 (l. 5-6), I. posits a lacuna after the Demosthenes quotation, Chiron before it; I.'s proposal (explained in her note d on p. 509) clearly seems the better one.
As this review has already gone on too long, and as I am not a native speaker of English, I will add only comparably few remarks about the translations, all of which come helpfully equipped with footnotes explaining various matters raised in the text. Each of the three translators evidently conceived his task somewhat differently: H. gives a whole new translation of the Poetics (he even states that he did not consult his own earlier one while writing it, adding that in this new one he aimed "to give a somewhat closer rendering than the earlier", p. 22), while R. -- although already being a well-established translator of On the Sublime in his own right -- has tried to keep as much as possible to Fyfe's earlier version ("I have tried not to tamper with it where it did not seem positively misleading", p. 156), and I., on the other hand, apparently was allowed more freedom to deviate from Rhys Roberts' former version, on which her own, however, still is "based" (she does not herself comment on her translation in her introduction). A short comparison of the present translations with their earlier counterparts seems to confirm these differences of approach:
In H.'s case, his more recent translation indeed markedly differs from the 1986 one (while some turns of phrases, of course, still sound similar) -- though I wouldn't always agree that the newer one is closer to the original Greek than the former. Right at the beginning, in 1447a19 he translates SUNH/QEIA with "knack" and in a20 TAI=S EI)RHME/NAIS TE/XNAIS with "all the poetic arts", while in 1986 he used the more accurate expressions "practice" and "all the arts mentioned above" (would "all the poetic arts" include -- in Aristotle's view -- AU)LHTIKH/ and KIQARISTIKH/, which, however, have clearly been "mentioned above"?). In 1447b15, PROSAGOREU/ONTES is rendered rather loosely as "this is not to classify", and in b19 MA=LLON H)\ POIHTH/N has completely dropped out (not in the former translation, where also FUSIOLO/GON in this line is better rendered as "natural philosopher" than as the modernistic "natural scientist" now in the Loeb volume). One could raise more quibbles like this,21 but in general H.'s translation seems reliable and readable.
R., in his part of the work, faithfully keeps to his task of "merely" revising Fyfe's older translation; in comparing the original and the reworked version, one can really see how respectfully he went about his job (which, however, must have been fully as laborious as writing an entirely new translation of his own), but everywhere he has introduced changes, the translation now is closer to the Greek or more comprehensible to our ears and minds. Quibbling, of course, is possible even here; sometimes one might wish that R. had introduced some more changes of his own.22
Compared to R.'s procedure, I.'s translation of On Style takes a much freer path with regard to its predecessor, of whom much less is still recognizable (a turn of phrase every few lines; very rarely a whole sentence has remained as it once was). But this much more extensive reworking (which makes this text really I.'s own translation) is all to the benefit of the modern reader. I. has consistently modernized terminology (rendering, e. g., KW=LON as "clause" instead of "member"), adapted sentence structure to the modern reader's expectations, and simultaneously managed to give a close rendering of the original (a closer one, in fact, than her predecessor). Again, criticism of details is possible23; but, again, this does not really detract from the overall achievement.
The book ends with three short indexes, each displaying the proper names and literary works cited in each of the three texts respectively. This is a useful addition; an even more useful one might have been a unified index listing all the Greek terms of literary criticism as they are employed by the three authors. In the predecessor volume, Rhys Roberts provided such an index (though of terms in English) at least for Demetrius; having something of the kind for all three texts together would have been a real treat.
All in all, this is a volume whose buyer gets real worth for his money: a generally reliable Poetics, a thoroughly re-thought On the Sublime and probably the best text of On Style currently available; so this book will be nothing less than essential for everyone doing work on ancient literary criticism.
 St. Halliwell, Aristotle's Poetics, London 1986; The Poetics of Aristotle: Translation and Commentary, London 1987; The Importance of Plato and Aristotle for Aesthetics, in: J. J. Cleary (ed.), Proceedings of the Boston Area Colloquium in Ancient Philosophy vol. 5, Lanham 1991, 321-348; "Aristotelian Mimesis Reevaluated", Journal of the History of Philosophy 28, 1990, 487-510.  D. A. Russell, 'Longinus' On the Sublime, Oxford 1964; Libellus de Sublimitate Dionysio Longino fere adscriptus, Oxford 1968.  In: D. A. Russell and M. Winterbottom (edd.), Ancient Literary Criticism: The Principal Texts in New Translations, Oxford 1972; reprinted in: Russell and Winterbottom (edd.), Classical Literary Criticism, Oxford 1985.  D. A. Russell, "Longinus Revisited", Mnemosyne 34, 1981, 72-86  D. C. Innes, Theophrastus and the Theory of Style, Rutgers University Studies in Classical Humanities II, New Brunswick 1985, 251-267; "Cicero on Tropes", Rhetorica 6, 1988, 307-325; Period and Colon: Theory and example in Demetrius and Longinus, Rutgers University Studies in Classical Humanities VI, New Brunswick 1994, 36-53.  R. was much more reluctant in admitting this possibility in his 1964 commentary (p. xxiii). The name Dionysius Longinus has recently been adopted, too, in the edition (together with Italian translation and commentary) by C. M. Mazzucchi, Milan 1992.  In p. 1448a15, H. adopts Vahlen's GA/R instead of the clearly corrupt GA=S; in 1448b36 it may be possible to write TA\ TH=S KWMW|DI/AS SXH/MATA (considering also 1149b3) instead of the singular; again, in 1450a16f., H.'s PRA/CEWS (reflecting the majority of the tradition) may be right as against Kassel's plural; in 1452b2, one might try to live with the transmitted E)/TI DE/ instead of adopting Vahlen's E)PEIDH/; in 1456a20, Castelvetro's TW=| QAUMASTW=| might indeed be a better antecedent for the following TOU=TO than the transmitted QAUMASTW=S; in 1457a4, Bywater's DH/ TOI/ (instead of H)/TOI) is conceivable, as is Vahlen's addition of KURI/WN before O)NOMA/TWN in 1458b16; in 1459b36, Kassel himself commented upon Twining's addition of TAU/TH| before H( DIHGHMATIKH\ MI/MHSIS as "fort. recte"; in 1460b18 one might consider doing without Vahlen's (admittedly very elegant) insertion A(/M' before A)/MFW.
In some cases, H. cites interesting conjectures missing in Kassel's apparatus: 1455a33 Tyrwhitt's insertion of MA=LLON before H)\ MANIKOU=; 1456a10 Vahlen's A)EI\ KRATEI=SQAI; in 1456a2, Bursian's conjecture H( A(PLH= (instead of the corrupt OHS), for which H. aptly compares 1459b9, probably deserves the place in H.'s text.
 In his translation of 1987, H. left out the second sentence as "probably spurious" (p. 37, 66).  In his translation of 1987, H. left out these three words.  One may also compare the remarkably contradictory translations of J. Vahlen, Beiträge zu Aristoteles' Poetik, Berlin 1914, 320 ("wenn einer ein A)/LOGON in den Mythos aufnimmt und es erscheint EU)LOGWTE/RWS, so muss auch ein A)/TOPON zulässig sein ...") and Bywater ("If the poet has taken such a plot, however, and one sees that he might have put it in a more probable form, he is guilty of absurdity as well as a fault of art") to see how slippery the ground is in this passage.  Here and in all following instances, I cite passages in On the Sublime and Demetrius by chapter and paragraph (or, in Demetrius' case, simply paragraph) and line-number counted from the beginning of the paragraph in this volume.  More convincing is Reiske's PEPEI=SQAI (considered by R. in a footnote), but the best solution might well be simple deletion; even if PEPAI=XQAI was the original reading of our sole manuscript witness (which is far from clear), it might have intruded there as a gloss on U(POTI/QESQAI.  R. writes "pages", but "folia" in the 1964 and 1968 editions.  In 1964, R. translated "a narrow aqueduct", in 1972 "an aqueduct".  It is probably no accident, that R.'s 1972 translation simply left them out.  Or leave out the comma both before and after KAQA/PER TA\ SW/MATA, as Mazzucchi does.  In 7.3 (l. 7) read MO/NHS instead of MO/HNS, in 8.2 (l. 8) POIHTH=| instead of POITH=|; in 8.4 (l. 5) the words OU)DE\N OU(/TWS W(S TO\ GENNAI=ON PA/QOS, E)/NQA XRH/ have dropped out between W(S and MEGALH/TORON; in 9.10 (l. 8) read U(=IAS instead of OI(=AS, in 10.4 (l. 1) PARAKOLOUQOU/NTWN instead of PARAKOKOUQ-, in 12.4 (l. 9) SKHPTW=| instead of SKHPRW=|, in 28.2 (l. 3) H(MI=N instead of U(MI=N, in 33.5 (l. 8) NH\ *DI/A instead of simply *DI/A, in 39.3 (l. 13) KAQISTA=SAN instead of PAQISTA=SAN, in 43.3 (l. 6) PROSKOSMHMA/TWN instead of PRO-, in 44.1 (l. 9) GI/NONTAI instead of GENNW=NTAI, in 44.5 (l. 4) E)GKEKLEISME/NWN instead E)GKEKLES-; in 44.8 (l. 3) the words EI)=NAI/ TINA LO/GON, A)LLA\ TOIOU/TWN E)N KU/KLW| have dropped out between U(STEROFHMI/AS and TELESIOURGEI=SQAI. The paragraph numbers 34.3 and 39.3 must be moved one line down; in 42.2 (l. 2) a colon has perobably dropped out after E)KTA/DHN.  Poetics: p. 23, l. 8 from bottom read Oxyrhynchus instead of Oxyryn-; p. 126 n. 3, Vahlen's supplement is not complete (the DI' after MIMH/SASQAI is missing). On Style: on p. 347, note "c" should be indicated two lines below its present position (after "metre"); in 15 (l. 7) TA/S before PUKNA/S has got a Latin "t"; on p. 384, the information in n. 2 has got wrongly turned around and must read: "E)/XOMEN edd.: E)XOME/NH P." (compare Chiron); in 58 (l. 3) delete LE/GOUSIN after E)PILE/GOUSIN. In n. 4 on p. 442 the second MIMIKW/TERA must be changed into MIMHTIKW/TERA; in 184 (l. 6f.) H)\ "DIATELEI= O(/LON TO\N BI/ON" has not been translated (see below n. 23); in 200 (l. 2) E)/STI PO/LIS has dropped out before *E)FU/RH.  Démétrios, Du Style, texte établi et traduit par P. Chiron, Paris 1993 (Collection Budé).  In a very few places I. might have considered some proposals made by Chiron: e.g. in 146 (l. 6) a lacuna after LAMPRO/TEROS, in 156 (l. 4) reading *NH/PIOS (suggested by Toup) instead of the corrupt E)PI/HS, in 184 (l. 7) reading E)CE/XEAS (proposed by Radermacher) instead of E)CE/XEIS. In 162 (l. 6) Chiron's conjecture KAI/TOI DIAFE/ROUSAI to save the almost certainly intrusive KAI/ TI DIAFE/ROUSI (deleted by Spengel, whom I. follows) is probably in vain.  Examples: in 1452b14f. "its formal and discrete sections" seems less easily understandable (and no more accurate) than "the quantitative divisions of the genre" in H.'s earlier translation of KATA\ DE\ TO\ POSO\N KAI\ EI)S A(\ DIAIREI=TAI KEXWRISME/NA. The somewhat enigmatic sentence PIQANW/TATOI GA\R A)PO\ TH=S AU)TH=S FU/SEWS OI( E)N TOI=S PA/QESI/N EI)SIN (enigmatic, because the transition from the preceding sentence seems not altogether clear) in 1455a30f. was formerly rendered by H. "assuming the same natural talent, the most convincing effect comes from those who actually put themselves in the emotions" and now "a natural affinity makes those in the grip of emotions the most convincing"; I prefer the former version. In 1457a2 and 9, H. translates E)PI\ TW=N A)/KRWN with "at the ends", but Margaret Hubbard -- in the 1972 volume cited above in n. 3 -- probably more correctly with "at either end".  E. g. in 1,3 (l. 6f.) he keeps Roberts' "this alone ... clothed them with immortal fame", a rather loose rendering of OU)K A)/LLOQEN H)\ E)NQE/NDE ... TAI=S E(AUTW=N PERIE/BALON EU)KLEI/AIS TO\N AI)W=NA; might one come closer to Greek by saying "by this alone ... they invested eternity with their fame"? In 2,1 (l. 6) the translation of FUSIKA\ E)/RGA is "works of natural genius", but R. in his own 1972 translation wrote (quite accurately) "natural products"; in 3,4 (l. 10f.) O)LISQAI/NOUSI D' EI)S TOU=TO TO\ GE/NOS is rendered by Roberts/R. with "Writers fall into this fault"; in 1972 R. kept closer to the imagery of the text by writing "Writers slip into it". In 7,1 (l. 12) the old/new version renders A)NAPTUTTO/MENA with "peeled off"; again, R. in 1972 is more exact ("dissected").  Examples: in 35 (l. 3) I.'s translation "he seems to limit the period not to two clauses but to three or more" does not do full justice to A)LLA\ KAI\ TRISI/ (here's one case, where Roberts' old rendering seems to be more accurate: "Archedemus seems ... to include three or a greater number"). Is "distinctive" really the most appropriate translation for E)CHLLAGME/NH in 77 (l. 2)? Chiron has "détourné de son sens propre", and in Poet. 1457b3, H. translates E)CHLLAGME/NON with "modification". In 112 (l. 4), "plagiarize" is a very harsh word (though sanctioned by LSJ) for META/QESIS (compare Chiron's "transposer"; one would like to know what Demetrius had in mind, when he chose this term to characterize Herodotus' use of the poets). More serious objections might be raised against I.'s translation of POI/HMA GA\R A)/KAIRON YUXRO/N, W(/SPER KAI/ ... in 118 (l. 3f.) with "A line of verse in prose is out of place, and as frigid as ..."; here, Roberts' old rendering "A bit of verse out of place is just as inartistic as ..." seems more accurate, if a bit stilted (compare also Chiron's translation). In 184 (l. 2f.), I. should not have translated DIATELEI= TO\N BI/ON O(/LON with the quite normal word-order "he passes his whole life", but perhaps with "passes his life entirely"; in the following sentence Demetrius' presentation of the more normal word-order DIATELEI= O(/LON TO\N BI/ON ("... his whole life") has unfortunately dropped out of the translation. E)NAGW/NIOS in 193 (l. 1) should perhaps not be rendered simply by "immediacy", but at least by "immediacy of debate" (I.'s own former translation; compare Chiron's "aux joutes oratoires"); in 226 (l. 9) U(POKRITH=| PRE/PEI is rendered rather vaguely with "suits oral delivery"; in 1972, I. had followed Roberts in translating "suits an actor", which is closer to the text.