The Brill Dictionary of Ancient Greek

1 Preamble

The publication of a new “major” lexicon of Ancient Greek is a significant event.¹ The new Brill dictionary (“ge”) is of the same size and scope as the lexicon of Liddell and Scott, the current standard work. Moreover, the new lexicon has been published complete in one volume, unlike the other major lexicon project of our time, dge, still progressing through publication in parts.² Though ge is based, it is true, on a previous lexicon, the completion of the project in just four years is a prodigious achievement. The co-ordination of the team that produced this result—there are at least 15 names mentioned in the preface—was a major undertaking in itself. Similarly, the Italian predecessor (“gi”) involved at least 60 collaborators under the leadership of Franco Montanari.³

The character and merits of the new lexicon are indicated in the preface to ge by the three leaders of the team that produced it.⁴ Though many statements are made and most are clear, some reading between the lines is required to understand the presuppositions of the editors and what this lexicon attempts to achieve. The editors naturally draw attention to the improvements (new evidence; new lemmata; corrections to gi; inclusion of proper names), but they also appear to recognise that ge is not, and could not possibly be, the equivalent of a completely revised lsj (cf. §8). They do claim that “this Greek-English project is presented as an enhancement of lexicography,” and that the dictionary “presents a critical approach to lexicography in and of itself,” statements that are difficult to interpret, if not meaningless. Two further assertions will be commented on in the next section.

It is of course impossible to assess the entire contents of a massive reference tool such as a lexicon: any review necessarily involves sampling. My samples and discussion will have in view three main themes or aspects: (a) the provenance of ge; (b) ge as compared with lsj; and (c) the current state of the lexicography of Ancient Greek. The samples are presented under various headings, but all will be found to illustrate these three topics.

2 Provenance of ge

Dependence on predecessors is a general phenomenon of lexicography. Rarely has a lexicon been created afresh, that is, on the basis of a new collection of data subjected to a new lexical analysis. Only if the compilers are prepared to spend up to a century can it be done. The Oxford English Dictionary was created in this way (70 years); so was the Oxford Latin Dictionary (50) and the recently completed Dictionary of Medieval Latin from British Sources (100). lsj was not; neither was ge. Both descend from a line of predecessors, which in ge’s case include lsj itself. The provenance of ge may be summarised as follows:

• ge 2015
• < gi (1st ed. 1995; 2nd ed. 2004); 3rd ed. 2013
• < Rocci (1st ed. 1939; 2nd ed. 1941); 3rd ed. 1943
• < lsj 9th ed. in 10 parts, 1925-1940; + Bailly 1894
• < ls 8th ed. 1897 …
• < ls 1st ed. 1843
• < Passow 4th ed. 1831
• < Schneider-Passow (1st ed. 1819-23); 3rd ed. 1827
• < Schneider (1st ed. 1797-98); 3rd ed. 1819

ge derives from gi, which was based on an Italian predecessor by Rocci,⁵ compiled on the basis of lsj, with occasional help from Bailly.⁶ So ge actually derives from lsj, via translation into Italian and back again into English. lsj’s own history goes back through a line of revisions to a German ancestor, the work of Passow; and Passow itself rests on earlier predecessors (of which enough).⁷ It cannot be said that any of these lexicons was a completely new work.

All this is acknowledged in Montanari’s preface to ge (and its predecessor, the Prefazione to gi), and in a further preface by Nagy, Muellner, and Goh.⁸ Despite this, the latter editors make a surprising admission:

When Brill asked the three of us in 2010 to undertake the editing of an English version of the Greek-English dictionary of Franco Montanari … we had not yet fully reckoned with the fact that the original “Liddell and Scott,” later transformed into “Liddell-Scott-Jones’ or “lsj” … cannot really be called an “original” Greek-English lexicon.

Furthermore, after summarising the history of lsj back to Schneider, they characterise the process as “cross-fertilization,” adding “we are convinced that such cross-fertilization is at the very core of lexicography.” Some “cross- fertilization” there may be, but this is hardly the term to describe what is in fact a chain of dependence. Still more surprising is this statement of the editors regarding their own lexicon:

[I]t is to be emphasized that the lexicon is not a translation of the Italian definitions in and of themselves.

Evidence to disprove this claim can be found on every page of the lexicon when set beside its predecessor gi.

Does any of this matter? It does, because not everything in the Liddell and Scott tradition is reliable. There is a strong possibility that at every stage older material has been simply taken on trust and never re-examined. Decisions about the lexical analysis, the definitions, and the selection of data were often made long ago and have not been questioned when they ought to have been. The revision process to which Liddell and Scott was subject guaranteed this result: for practical reasons most of the material was carried on unchanged in each edition and a thorough revision could not be undertaken. Supplementary material added piecemeal each time often did no more than create confusion. Most users of the great lexicon—and even compilers of new lexicons—are unaware of these problems, though there has been no lack of remark on them.⁹ A first example to illustrate the point:

µεταδετέον

ge: µεταδετέον [µετά, δέω] verb. adj. it is necessary to unbind Xen. Hip. 4.4.

gi: µεταδετέον [µ., δέω] vb. bisogna slegare Xen. Hip. 4.4.

Rocci³ (1943): µετα-δετέον, vrb. da δέω, si deve sciogliere, ἀπὸ τῆς φάτνης, dalla mangiatoia, Sen. Eq. 4, 4.

lsj Revised Suppl. (1996): µεταδετέον, for ‘one must untie’ read ‘one must change the tethering’ (from one place to another)’

lsj: µεταδετέον, one must untie, X.Eq. 4.4.

ls⁸ (1897): µεταδετέον, verb. Adj. one must untie, Xen. Eq. 4, 4.

ls¹ (1843): µεταδετέον, verb. Adj., one must untie, Xen. Eq. 4, 4: from µεταδέω, δήσω, to tie differently; to untie.

Passow⁴ (1831): µεταδέω, δήσω, (δέω, Irr.) umbinden, d. i. anders od. anderswohin binden.

In this case a mistake has been carried on undetected all the way from the first edition of ls to ge. The meaning “untie” is simply wrong for µεταδέω: it appears to have arisen from a mistranslation of umbinden as unbinden. The correction in the Supplement was missed and the mistake sailed on.¹⁰ This is a trivial example—or maybe not, to those interested in ancient horse management—but it illustrates the phenomenon.¹¹ The next case is more serious.

κάµινος

ge: κάµινος … oven or furnace, forge … || flue for heating a room … || alcove vt Num. 25.8.

gi: κάµινος … forno o fornace, forgia … || condotto per riscaldare una stanza … || alcova vt. Num. 25.8.

Rocci: κάµινος … forno; fornace; fucina … alcova, Set. Num. 25, 8.

lsj: κάµινος … oven, furnace, or kiln … alcove, Lxx Nu. 25.8.

The meaning “alcove” for Num. 25:8 has descended to ge from lsj. It has no connexion with the known meanings of κάµινος or justification in the context of Num 25:8. What is the evidence for it? Investigation reveals that it is an alternative rendering offered in the margin of the Revised Version, that is, an English rendering of the Hebrew word translated by κάµινος at Num 25:8.¹² The underlying Hebrew is not a certain guide to the meaning of the Greek word that translates it; a guess at the meaning of the Hebrew in an English translation is no guide at all. By what process this proposal came into lsj we do not know, but it has no right to be there.¹³ The evidence for the meaning of κάµινος is to be sought in Greek, in usage contemporary with the lxx, if available. It happens that we have good material in documents of the third century bc. These not only show κάµινος in its usual meaning “furnace, kiln, oven,” but also reveal that κάµινοι in Ptolemaic Egypt could be sizable structures used for other purposes, such as storage of wine and the keeping of animals.¹⁴ From this it becomes clear that a κάµινος could have provided just the right secluded space for the illicit activity described in Num 25:8.¹⁵

3 Structure of Verb Entries

Let us take a straightforward common verb for a view of the general structure of an entry in ge and a comparison of ge with lsj.

κλαίω

ge: κλαίω … impf. ἔκλαιον … || fut. κλαιήσω …; mid. κλαύσοµαι … 1 act. A to weep, lament Il. 1.362 … B to suffer ill-treatment, undergo a punsihment [sic]: κλάοις ἄν, εἰ ψαύσειας you will be punished, if you touch them Aeschl. Suppl. 952 … C to weep for, bewail, grieve over > with acc. Il. 20.210 Od. 1.363 … || to call by crying: κλάειν µάµµας καὶ τιτθάς to cry for mummies and nurses Arr. … 2 mid. A to weep, lament Il. 18.340 … B to suffer, undergo punishment or harm, usually fut. Aristoph. Ve. 1327 etc.; … C to weep for, bewail > with acc.… 3 pass. to be bewailed Aeschl. Ch. 687 … | impers. µάτην ἐµοὶ κεκλαύσεται … Aristoph. Nub. 1436.…

lsj: κλαίω … i. intr., cry, wail, lament, of any loud expression of pain or sorrow, … Od. 10.201 … 2. αὐτὸν κλαίοντα ἀφήσω I shall send him home crying, howling, i.e. well beaten, Il.2.263: freq. in Att., κλαύσεται he shall howl, i.e. he shall suffer for it, Ar. V.1327 … ii. trans., weep for, lament … Od.1.363 … :—Pass., to be mourned or lamented …: impers., µάτην ἐµοὶ κεκλαύσεται … Ar.Nu.1436. 2. cry for, of infants, µάµµας καὶ τιτθάς Arr. … iii. Med., bewail oneself, weep aloud, A.Th. 920 … 2. trans. bewail to oneself.… S.Tr….

The general layout in ge is very similar to lsj’s, with the usual tumble of glosses, divided by a numbering system that requires concentration. ge gains in clarity by using bold for the meanings; but its numbering system is harder to pick up and somewhat counter-intuitive. The letters A, B, etc. (enclosed in boxes) are subordinate to the numbers 1, 2, etc. (enclosed in black blobs). While the blobs catch the eye, the numbers in them are very small; the boxed letters are quite hard to spot. In addition, there are subdivisions introduced by double and single bars (||, |).¹⁶ The explanation of these given in the endpapers is clear with regard to marking tenses and moods in the “principal parts section” of the entry (as in “|| fut. κλαιήσω” above), but is much less clear about why some senses are introduced in this way: the unsatisfying explanation given is that they are “further subdivisions of the entries.” In the case of κλαίω above, it looks rather as if “|| to call by crying” ought to have been another numbered sense.

These are somewhat minor matters. More importantly, we see in κλαίω a feature carried through systematically in all verb entries, a division into active, middle, and passive voices (marked by numbers 1, 2, 3). This is a significant innovation, introduced by Montanari in gi. While it may be helpful to the beginner, and even to some advanced scholars, it may not be a wise one from the point of view of lexical analysis. It has the effect of changing the upper level of the analysis from semantic to morphological; that is, lexical meaning is subordinated to the active-middle-passive distinction. The result is that in verbs (though only verbs) the primary divisions into 1, 2, and 3 do not indicate different senses but different voices, and consequently the same meanings may appear, indeed must often appear, twice or even three times in the entry. So in κλαίω, the meanings in 1 act. A, B, C are the same as those in 2 mid. A, B, C (“weep,” “suffer,” “weep for”); and sense 3 pass. gives a meaning which is just the passive of 1C and 2C (“weep for”).

This innovation can also lead to trouble of a different kind in the many Greek verbs in which form and meaning do not necessarily match.¹⁷ Examples: (a) λαµβάνω is normally middle in form in the future (λήψοµαι), but has an “active” meaning; so ge has to present this as “1 act. (with fut. mid.).” (b) The usual aorist of βούλοµαι is passive in form but not in meaning; ge gives up the division by form for this verb and simply says, in the principal parts section, “aor. only pass. ἐβουλήθην or ἠβ-.” (c) The verb βιάζοµαι has the annoying characteristic of being usable in the present middle form in both an “active” and a passive sense; ge’s solution is “1 act.… 2 usu. mid.… 3 pass.”; but this means that 2 mixes form and meaning.

As for verbs that have suppletives with different voices, major difficulties are unavoidable. ge in some cases simply gives up the division by voice, as in ἔρχοµαι and its aorist ἦλθον, which are wrapped together; in others various compromises are required.¹⁸ In ἐσθίω, all these problems come home to roost. The entry is simply divided into “1 act.… 2 pass.” Under the first, the active ἔδω is mentioned, but we are not told that the usual future of ἐσθίω is actually the middle ἔδοµαι (in Class., but later φάγοµαι); the standard aorist ἔφαγον, formed from a different root, is also not mentioned: one has to know that there is a separate entry for ἔφαγον, where its relation to ἐσθίω is noted; and although coverage of later Greek is said to be a feature of this lexicon, nothing at all is said about τρώγω, which came in as the present to replace ἐσθίω (cf. below). One would not want to suggest that lsj has solved all these problems, but its presentation of the voices and suppletives is generally much clearer than ge’s, and its entry for ἐσθίω covers everything with great efficiency.

All in all, as regards layout and convenience for the user, there is not much to choose between ge and lsj. Both present us with a packed entry offering a range of possible meanings, arranged in sense-divisions that may convey little to most readers. The selection of examples is overloaded and mostly from Classical authors. We have become used to finding our way through this in lsj; we will have to do the same with ge. How are such entries actually used? I suspect that most readers simply pluck out from the offerings a gloss that seems to fit the passage they are reading; whether it is the right one is often a matter of chance (unless of course they can locate a citation of the passage).

Let us follow up τρώγω, just mentioned in relation to ἐσθίω, and see how it has fared:

τρώγω

ge: τρώγω … 1 act. to gnaw on, nibble at, of animals: τ. ἄγρωστιν to nibble on weeds Od. 6.90; … | of sick people … || to eat (raw), chew on, of pers., esp. vegetables and sweets … || later simpl. to eat (= ἐσθίω) … 2 pass. to be eaten: τρώγεται ἁπαλὰ ταῦτα καὶ αὖα the fresh and dry are eaten Hdt. 2.92.4; …

lsj: τρώγω … [i.] gnaw, nibble, munch, esp. of herbivorous animals, as mules, τ. ἄγρωστιν Od.6.90; of swine … of cattle … of human beings in disease … ii. of men, eat vegetables or fruit … of dessert, eat fruits, as figs, almonds, etc.… of small fish as hors-d’oeuvres … iii. later, simply eat, serving as pres. to ἔφαγον instead of ἐσθίω …

The basic idea of this verb is “gnaw, nibble,” as stated, an action typical of animals but readily extended to humans. But it is rather the kind of eating, not who is eating or what is being eaten, that characterises τρώγω and distinguishes it from ἐσθίω, as Chadwick pointed out.¹⁹ lsj’s division into i., applied to animals and ii., applied to humans (with leakage in “of human beings …” in i.) is not necessary and the two could have been collapsed into one lexical meaning. Later on, from about i ad, by the same process by which Lat. manducare (“chew”) developed into Fr. manger (“eat”), this verb came to mean simply “eat,” making it equivalent to ἐσθίω, as lsj correctly report.

In ge the distinction between animals and humans is kept, with the meaning stated more or less satisfactorily (though “to eat (raw)” is not a good idea); but the insertion of the later meaning “eat” into this section, with only || to introduce it, underplays its significance as a distinct development in meaning, which leads in the long run to the Modern Greek word for “eat.” Then when the active-passive distinction is imposed, it creates problems. It looks to the user as if the passive voice is only used with the (sub-) meaning “eat,” a meaning said to be “later” (in 1.), yet cited from Herodotus (in 2.). In fact this example is the passive of “gnaw, nibble, chew” not “eat”: Hdt. 2.92.4 τρώγεται δὲ καὶ ἁπαλὰ ταῦτα καὶ αὖα = “and these [edible kernels] are nibbled both fresh and dried.”²⁰ In short, ge’s entry, though ultimately derived from lsj’s, has lost more than it has gained in the process of revision and translation.

4 Lexical Analysis

The core of the lexicographer’s task is to state the meanings of words, that is, to analyse the available data for each word into its “lexical meanings,” with suitable definitions. This is a very challenging exercise and good results are not guaranteed; they are usually worse when an entry has passed through many hands in its history. The samples in the previous section have already illustrated problems in ge (especially τρώγω), and others will be found in later sections (notably ὀψάριον, §5). Here I explore a typical noun frequently attested and with a range of meanings.

5 Documentary Evidence

The main purpose of lexicons is to supply meanings, but we also consult them for other information, especially attestation. Innumerable discussions in the scholarly literature offer statements like “first found in …,” “common in …”. On what are they based? On a lexicon usually, and usually lsj. Lexicons thus have an important role to play in this regard. Yet it cannot be said that the present state of play is satisfactory. In the lsj tradition the attestation or evidence for each word has accumulated unsystematically over centuries. It is primarily employed, as is right, to establish and illustrate the meanings, with the additional aim of supporting the reading of standard texts. But there is no guarantee that the picture given is a reliable guide to the spread and currency of the word. Until there is a systematic gathering of evidence, with some sort of statistical information, this situation will continue. We cannot expect ge to have solved this problem, but we would expect at least to see no loss of existing data, and the addition of new data where significant. Let us take as a test documentary evidence, i.e., inscriptions and papyri, of which it is said gi/ge “makes substantial use.”²⁶

εὐίλατος

ge: εὐίλατος -ον [εὖ, ἱλάοµαι] kind, favorable, merciful with dat. vt Ps. 98.8 etc.

gi: εὐίλατος -ον [εὖ, ἱλάοµαι] benigno, propizio, misericordioso, con dat. vt. Ps. 98.8 ecc.

Rocci: εὐ-ίλατος, ον [ῑ, ἱλάοµαι] propizio; pietoso, I.; Set.

lsj: εὐίλατ-ος [ῑ], ον, (ἵληµι) very merciful, of deities, PCair.Zen. 34.19 (iii bc), ig … gdi … etc.; … Lxx Ps….; also … PPetr….; later written … gdi … upz 109.6 (i bc).

The documentary evidence for this word, mainly in inscriptions, is extensive and important, showing as it does that the lxx was using a current term and how it was used.²⁷ lsj had quite a good collection. What has happened to it in gi/ge? It has disappeared. Rocci at least had “I.” (= “Iscrizioni”) to alert the reader; gi let it go. Hence in ge it looks as if εὐίλατος is a word used only in the lxx.

ἀποσκευή

ge: ἀποσκευή … A baggage … || household goods vt Gen. 14.12, al.… | all members of the house vt Exod. 10.24 | the whole people, aside from the adult males vt Exod. 12.37 | in the army attendants vt Iud. 7.2 || excrement, filth Strab. … B suppression, elimination Ios. a.i. 18.2.4(41).

lsj Revised Suppl. (1996): … add ‘2 soldier’s encumbrances, i.e. family, PBaden 48.9 (ii bc), upz 110.199 (ii bc); dependants, Lxx Ge. 46.5, al.’

ge’s analysis is very odd to begin with; but the main point is that no documentary evidence is mentioned, even though it is important for the history of the word and indispensible to determining the meaning in the lxx. lsj Supplement supplied some, enough to show the new meaning family, dependants that appears in Ptolemaic Greek.²⁸ Not only ought some of this evidence to have been reported in ge but even a brief look at it would have greatly improved ge’s account of the lxx uses.²⁹

ὀψάριον

ge: ὀψάριον -ου, τό [ὄψον] course of a meal, a little dish, esp. of fish Aristoph. fr. 45 Plat.¹ 102.2 Pherecr. 32 etc.; ἔχει πέντε ἄρτους κριθίνους καὶ δύο ὀψάρια he has five barley loaves and two fishes nt John 6.9.

lsj: ὀψ-άριον, τό, Dim. of ὄψον, Ar.Fr.45, Pl.Com.95, Pherecr.27, Philem.32, Test.Epict.6.11, PPetr. 3 p.327 (iii bc), PCair.Zen.440.3 (iii bc), etc.; … a jar of pickled fish, BGU1095.17 (i ad), cf. PRyl.229.21 (i ad), Ev.Jo.6.9, al., OGI484.12 (Pergam., ii ad).

lsj Revised Suppl. (1996): ὀψάριον, after ‘dim. of ὄψον’ insert ‘foodstuff, esp. fish’ and add … seg 26.382 (Athens)’.

Rocci: ὀψάριον, ου, τό, dim. di ὄψον, propr. piccolo companatico, ma, in Atene, spec. pesciolino; pesce, vari Com. in At. 385; P.; δύο ὀψάρια,Volg. duos pisces nt. Io. 6, 9.

As well as a complete lack of documentary evidence in ge, we find a very unsatisfactory statement of meaning. lsj had good data to work with (added in lsj⁹), but committed the sin of not telling us the meaning, instead relying on “Dim. of ὄψον” to do the job. This is worse than useless: the user is given the impression that ὀψάριον is a small ὄψον—whatever that is—and turns to ὄψον only to see a range of meanings, any of which could be relevant. lsj Supplement tried to remedy this but didn’t do it well: “foodstuff” covers all food; “esp. fish” vacillates. ge has made it worse.³⁰ Rocci was good within his limits. What is needed, instead of guesses based on a preceding lexicon, is a fresh assessment of all the data and a clear statement of the meaning so far as it can be known.

These three examples barely touch on the amount of under-reporting of documentary evidence in ge. In regard to the papyri alone, the extent of it is dumbfounding given the claims made. Some are standard and frequent technical terms, such as ἔντευξις (“petition”). Here are 50 well-attested words in papyri that lack any reference to that evidence in ge:

ἀµάω, ἄµφοδον, ἀξίνη, ἀξιόχρεως, βῆµα, διαθήκη, διάπρασις, διαπράσσω, διαστέλλω, διάστηµα, διαστολή, διαστολικόν (no entry), διασῴζω, διαταγή, διάταγµα, διάταξις, διατάσσω, ἐγγύη, ἐγγυητής, ἔγγυος, ἐλαιών, ἔντευξις, ἐντυγχάνω, ἐξακολουθέω, ἐπικεφάλαιον, ζωή, ζῷον, ἡµίονος, κάµινος, κοινωνός, κόκκος, λογίζοµαι, µέρος, µεταβολή, µοναστήριον, ναύκληρος, νοµός, ὀφείληµα, παράδεισος, παραθήκη, πράκτωρ, πρόγραµµα, προσευχή, προστάσσω, πυρός, ῥόα, σιτοµέτρης, ὑποζύγιον, φερνή, χωρίον.

Of course this simply reflects the reality, that to try to graft onto a lexicon of Classical Greek adequate coverage of even a part of the vast body of later Greek is impossible without a giant effort.

6 nt and lxx

Coverage of the nt and lxx in ge is difficult to assess fully (and fairly). It is evident that references to these texts are incorporated frequently throughout gi/ge (as they were in Rocci’s Vocabolario).³¹ Moreover, much of the nt/lxx vocabulary is no different from that of other ancient texts, so most words and meanings are as it were automatically covered. Even so, ge comes nowhere near providing the in-depth treatment available in the specialist lexicons,³² and in some cases there are significant gaps. This is the natural outcome of a tradition in which the nt and lxx were treated as an adjunct to the lexicon of Classical Greek. The coverage was never really satisfactory in lsj or any of its predecessors, even though these texts received some notice from the first edition onwards. ge’s material has come into existence in a similar way and the results are similar.

7 A Particle

Should we expect lexicons to help with particles? Whether we should or not, we certainly do. We hope to get a quick indication to help us out. Few of us are ready to face the ordeal of reading Denniston’s famous book or some other specialist study (inaccessible to us anyway). The lexicons do try to cover particles, usually in some depth. In ge µέν, for instance, is given a very extensive treatment filling three columns; the material in lsj is of similar length; Rocci was shorter but still thorough. We have no reason not to trust most of this. Rather than try to examine µέν, let us take a less frequent particle, καίτοι, to see (a) whether ge is an advance on lsj; and (b) how well both lexicons cover the data and usage of καίτοι.

καίτοι

ge: καίτοι or καί τοι adv. certainly, truly, indeed: κ. ἐµοί yes, for me also Il. 13.267 cf. Eur. Med. 344; καίτοι σοφοῦ παρὰ φωτὸς εἰρηµένον indeed said by a wise man Sim. 37.12 || advers. however, yet, nonetheless: κ. τί φηµι and yet what am I saying Aeschl. Pr. 101; καίτοι γε nevertheless Xen. Mem. 3.12.7 | concess. although, even if: καίτοι περ although Hdt. 8.53.

Rocci: καίτοι, certo; a dire il vero; veramente, κ. ἐµοί, certo, anche a me, Il.13, 267; v. 1, 426; Eu. Med. 344.—b) tuttavia; pure; per altro; nondimeno, Att.; κ. τί φηµι; ma che dico? Eschl. Pr. 101; καίτοι γε, Sen. Mem. 3, 12, 7.—c) quantunque; sebbene, Att.: κ. περ, quantunque, Er. 8, 53.

Bailly: καίτοι, adv. 1 et certes, et en vérite [sic], Il. 13,267; Isocr. Pan. 67, etc. || 2 quoi qu’il en soit, cependant, toutefois, Att.; … κ. τί φηµι; Eschl. Pr. 101; … || 3 quoique, devant un part.… renforcé par περ. καίτοι περ, Hdt. 8,53.

lsj: καί τοι, and indeed, and further, freq. in Hom. with one or more words between, Il. 1.426, al.; καὶ σύ τοι E.Med.344; καὶ τἆλλά τοι X.Cyr.7.3.10: once in Hom. as one word, Il.13.267. ii. after Hom. usu., and yet, to mark an objection introduced by the speaker himself, freq. in Rhetorical questions, καίτοι τί φηµι; A.Pr.101 …: without a question …: strengthd., καίτοι γ᾽ … etc.: mostly separated, καίτοι … γε … X.Mem.3.12.17, etc….; so καίτοι περ v.l. in Hdt. 8.53. iii. with a participle, much like καίπερ, Simon. 5.9 …

It will give us some perspective to start from outside the lexicons. Denniston’s analysis finds four main uses of καίτοι (adversative; continuative; logical; combined with other particles), with subgroups in some, a total of 14 sections. Each is illustrated, in Denniston’s usual fashion, by a wealth of examples, all from the Classical period except one. At a rough count he gives 250 examples.³⁹ Blomqvist covers the same ground for his selection of Hellenistic authors (330-30 bc). He includes a table of frequency in his texts and some Classical authors. It shows that while there are fewer in later texts (70+), the numbers in Classical authors add up to hundreds; these include high numbers in prose writers such as Demosthenes (190+), Plato (120+), and Aristotle (200+).⁴⁰

It is obvious that our lexicons offer an exceedingly limited selection of evidence for καίτοι. ge cites only six examples. Only two of these are from prose (Xen., Hdt.) and none are post-Classical. lsj has more, a total of 27 (with “etc.” sometimes), and they are slightly more representative, including more from prose (D., Pl.) and even a couple of late authors (Plb., Philo). But the coverage is still very inadequate in lsj, and in ge entirely unsatisfactory.

lsj appears to recognise three meanings or uses of καίτοι. The second and third are clearly described, but the first is rather strange and unclear: it seems as if and indeed, and further is frequent in Homer, but only if καί τοι is split by another word (except once), and that this use occurs only a couple of times later. The fit to the noted examples does not seem right either: for instance, the first, Il. 1.426 καὶ τότ᾽ ἔπειτά τοι εἶµι, (“and then after that I will go”), can hardly be an example of καί + τοι. Close scrutiny suggests the first meaning was never intended to be a separate sense back in Passow and ls¹ where it originated, but was an ambient meaning to cover the word as a whole; treatment of Homer was simply the next item in Passow. Proof of this is that sense i. was originally and yet, yet in ls¹ (< Passow doch, und doch, doch auch, dennoch). It was changed in the 5th edition (1861) to and indeed, and further, the meaning we have now, at the same time as a new sense ii. and yet was introduced. (lsj’s sense iii., the concessive use, was separated from the beginning.) These changes caused confusion in sense i. The match between these apparent three meanings in lsj and Denniston’s analysis is a worry too. Sense ii. matches D.’s (1) “adversative” use, and iii. matches (1.vii) the “concessive” use; but where are D.’s (2) “continuative” and (3) “logical” uses? Various other problems could be pointed out but need not be gone into here.

ge’s entry is no more than an attenuated remnant of its predecessors. The first sense or use (there are no numbers, only || and |) has been freshened up with input from Bailly via Rocci, but the illustrations doubtfully support it (one, Sim. 37.12, has migrated from sense 3).⁴¹ The second sense, again from Bailly via Rocci, is an inexact catch-all, and the translation of Xen. Mem. 3.12.7 (included in lsj only to illustrate καίτοι separated from γε) is simply wrong. The third is real but the example is a dubious v.l. (many others were available).⁴²

This look into καίτοι has enabled us to see not just what is wrong with the current entries but how their history has controlled the outcomes. The major underlying problem is reliance on an existing lexicon entry as the basis for the next, however much “revised.” Unless the old entry is completely laid bare and all its data are reassessed, its problems will be passed on or made worse. Hence our conclusion that (a) ge’s entry for καίτοι is not an advance on lsj’s; and (b) the coverage of data and usage is inadequate in both lexicons, but much worse in ge than in lsj.

8 Conclusions

The foregoing samples may be a small selection but they are representative of widespread phenomena in the new lexicon. They justify an assessment as follows. (a) Is ge a “better” lexicon than lsj? No. (b) Is it a competitor to lsj? Yes; but only because it is a major lexicon aimed at covering the same territory. (c) Is it the equal of lsj? No. (d) Can it replace lsj? No. (e) Will it serve the purposes of the ordinary student beyond elementary level? Yes, up to a point. (f) Can it serve the purposes of scholars? Again, up to a point. (g) Does it apply an improved method of defining meaning? No. (h) Is it better at lexical analysis than lsj? No. (i) Does it offer generally better coverage of meanings and attestation than lsj? No. (j) Does it actually provide better coverage of post-Classical Greek, as it claims to do? No, very seldom. (k) Does it omit information that lsj offers? Yes, often. (l) Is its presentation easier to follow than lsj’s? To a slight extent only. (m) Is its numbering system easy to follow? No. (n) What does it offer that lsj does not? Limited additions to the attestation of words and some meanings. (o) Is it basically a translation of its predecessor gi? Yes. (p) Is gi basically a revision of a predecessor? Yes.

ge is a useful lexicon, but it is not the great advance that we long for. The shape of the lexicography of Ancient Greek has not been changed by ge, but remains where it was. ge simply supplies another lexicon of similar size and coverage to lsj, doing the job no better and often rather worse. The principal reason is that ge is a descendant of lsj itself, with some revisions, but with most of the faults as well as merits of lsj preserved. The major, far-reaching overhaul that lsj’s material needs has not been undertaken, either by the editors of ge or by anyone.

Montanari rightly describes the situation in his preface to ge:

The Liddell-Scott-Jones dictionary is used by scholars of antiquity as the reference dictionary for Greek, although the need for a profound revision of lsj itself or for a completely new dictionary of Ancient Greek has been put forward many times and is indeed increasingly felt, in order to take into account more organically the advances of knowledge achieved over the years.⁴³

I read these words as a true statement of where we are, not a claim that gi/ge is itself the “profound revision” that is needed. Montanari wisely refrains from making such a claim. If after all he intended to imply it, or others take his words in that way, it needs to be stated that there is no justification for it.⁴⁴

What, then, is to be done? Lexicons like ge are a legacy of the past. They deal primarily with Classical Greek, and contain both too much and too little. They retain old material and have difficulty adding new. They also preserve the mistakes of their predecessors and have difficulty removing them. The coverage of post-Classical Greek is out of control because there is too much material. Trying to graft small portions of it onto existing lexicons is unworkable. Lexicons like ge are now a holding operation. Rather than continue to produce new lexicons based on old ones, the next generation of Greek lexicographers must rethink how to gather and present their material. Gathering of data by electronic means is now the only option; and a reappraisal of the meanings is essential. The whole corpus of Ancient Greek is too vast to deal with at once, but a systematic approach could move us forward. If the corpus were divided into manageable segments, the data could be gradually combined into one major electronic resource. Printed lexicons would still be needed, but could be slimmer and more selective, and based on better data.