CUNNINGHAM, IAN C., Harpocration and the "Synagoge" [Greek] , Greek, Roman and Byzantine Studies, 27:2 (1986:Summer) p.205 Harpocration and the LVval'wY11 Ian C. Cunningham 1. The Problem The only surviving work of Valerius Harpocration, Alexandrian PT,TWpl or ypaJ.L/-UXT/,KOfi2 of the later second centurY,3 is the AegLKov TWV 8EKa Pl1TOpwv, an alphabetical elucidation of words and phrases and especially names of persons and places in the canonical Attic orators.4 The full form of this text seems not to have been known in Byzantium till the thirteenth century.5 But an epitome, perhaps made in late antiquity, was available. It is independently extant in three manuscripts: E of the thirteenth century, D of 1496, and T (a copy of E) of ca 1540.6 Much earlier than any of these was the copy or copies used by one or more expanders of the Lvvaywyry AEgewv XPl1uiJ.LWv. This protean LvvaywyT, (hereafter L) cannot be ascribed to any one compiler or to an exact period, but it belongs to the revival of interest in scholarly matters commonly known as the first Byzantine renaissance, i.e., to the very end of the eighth or the first half of the ninth century.7 The terminus ante quem is the date of the lexicon of Photius, but that is by no means certain; 830 to 850 are the probable limits.8 L began as an expanded version of the lexicon of Cyril, an elementary compilation of glosses on the Bible, Homer, Euripides, and other well-known texts, probably by the fifth-century archbishop of Alexandria.9 In this form it exists in Paris eOis/.gr. 347 (A, 9th or 1 Suda A 4014. 2 Heading in MSS. D and E (see infra). 3 E. G. Turner, JEA 38 (1952) 92; B. Hemmerdinger, REG 72 (1959) 107-09. 4 Edited by I. Bekker (Berlin 1833) and by W. Dindorf (Oxford 1853); citations are as is customary by Dindorf's page and line, but the text and apparatus of both editions have been utilised. An edition was planned by Georg Wentzel, whose material was available for the RE article by H. Schultz (VII.2 [1912] 2412-16). 5 J. J. Keaney, TAPA 100 (1961) 201-07. 6 The variations from the full text can be seen in Dindorf's first apparatus. The substantial agreement of all the sources in selection of glosses, wording of abridge­ ments, and textual details shows that one epitome is in question. 7 Cf. P. Lemerle, Le premier humanisme byzantin (Paris 1971). 8 For a convenient summary of the evidence and literature see K. Alpers, Das attizis­ tische Lexikon des Oros (Berlin 1981) 72 and n.23. 9 Unpublished. Cf. A. B. Drachmann, Die Uberlie/erung des Cyril/glossars (Copen­ hagen 1936), with edition of the sections {3a, ()cr()E, Aa-AE. A version of this lexicon was incorporated into Hesychius; much information is given in K. Laue, ed., Hesychii 205 CUNNINGHAM, IAN C., Harpocration and the "Synagoge" [Greek] , Greek, Roman and Byzantine Studies, 27:2 (1986:Summer) p.205 206 HARPOCRATION AND THE LYNArilrH 10th cent.) and, apart from the letter a and a few additions else­ where, in Paris COisl.gr. 345 (B, 10th cent.) , and also with later interpolations in Berlin gr.qu. 13 (C, 15th cent.; now in Cracow, Bibljoteka Jagiellonska).lO It later underwent great expansion in sev­ eral stages, by which glosses from the following sources were incor­ porated: Apollonius' Lexicon Homericum, fuller versions of the rhe­ torical lexica printed as the fourth and fifth items in Bekker's Anec­ data graeca 1,11 glosses on Plato from Timaeus and elsewhere, Atticis­ tic lexica (Aelius Dionysius, Pausanias, Phrynichus, the Antiatticist, perhaps Orus) , Diogenianus, and Harpocration. It is not to be sup­ posed that the original ancient works were utilised, but rather as with Harpocration epitomes, extracts, and compilations, and these prob­ ably gathered in only a few manuscripts.12 Each of our three sources for this "erweiterte L" (as Wentzel called it)-B,13 the Suda 14 and Photius15 - has a different selection of these additions: the principal groupings are B Suda Phot., Suda Phot., and B Phot. The most eco­ nomical explanation of this is that there were three stages of expan­ sion, L', L", and L"', the two latter being based on the first; Suda used L", Bused L"', and Phot. used both. This is clearer in the diagram at the top of page 207.16 The much smaller number of coincidences between Band Suda are to be explained by omission in Photius, either by Photius himself or (more likely) in the course of the transmission of his lexicon, as the Alexandrini Lexicon (Copenhagen 1953-). An edition of Cyril was being prepared by the late Mark Naoumides. 10 a is printed from A by C. BOYSEN, Lexici Segueriani ~vlJa'YW'Yl1 A.E~EWIJ XP'YI(J'if..UJJ1J II inscripti pars prima (Marburg 1891-92, reprinted in Lexica !J!J!eca_minora [Hildesheim 1965] 12-60 [hereafter 'Boysen')), the remainder from B by L.IfACHMANN, Anecdota graeca I (Leipzig 1828 [hereafter 'Bach.')) 178-422. 11 Berlin 1814. These are entitled, respectively, ducwv OVO#UlTa [hereafter 'Bekk. IV'] and AE~EL~ P'YITopucai ['Bekk. v']. 12 Such as a precursor of B, which contains, in addition to ~, Apollonius, Phrynichus, Timaeus, the Antiatticist, and Bekker IV and v. 13 a in the manuscript B. 14 Suidae lexicon, A. Adler, ed. (Leipzig 1928-38). 15 Photii patriarchae lexicon, C. Theodoridis, ed. (Berlin 1982- ). 16 This is a very brief summary of a complex matter, intended only as essential back­ ground. Less important manifestations of ~, as in Paris Suppl.gr. 1243 and the scholia on Plato and Lucian, and controversial ones, such as the P'YITOPUCOV used in the Ety­ mologicum genuinum, are ignored. For further details see G. Wentzel, reviewing Boy­ sen's edition, GGA 155 (1893) 27-46, and "Beitrage zur Geschichte der griechischen Lexikographie," SitzBerlin 1895, 477-87 (reprinted in Lexica graeca minora 1-11); R. Reitzenstein, Der Anfang des Lexikons des Photios (Leipzig 1907) xxix-liii; K. Laue, Hermes 50 (1915) 376; A. Adler, RE IVA.l (1931) 675-717 S.v. "Suidas"; H. Erbse, Untersuchungen zu den attizistischen Lexika (Berlin 1950) 22-34; K. Alpers (supra n.8) 56-79; Theodoridis (supra n.15) I xxxv-Ix, lxxii-lxxvi, and GGA 235 (1983) 189-209. CUNNINGHAM, IAN C., Harpocration and the "Synagoge" [Greek] , Greek, Roman and Byzantine Studies, 27:2 (1986:Summer) p.205 IAN C. CUNNINGHAM 207 t I /t'", til til' / ""/ '" Suda Phot. B17 manuscripts are relatively late and have been subjected to abbrevia­ tion.Is The possibility that glosses that now appear only in one of them came from one or other forms of 12 and were omitted in the rest of the tradition is frequently present but can rarely be proved. All three also had other sources. It is obvious that only in a, where B is available, can this process be proved. Elsewhere 12" and 12'" can be separated only by analogical extension of the results obtained for a. Different views have been held as to the stage (s) at which Harpo­ cration was added to L. Wentzel, writing before most of a in Photius was known, included Harpocration among the common additions,19 Reitzenstein, the first to have a clear view of the stages of expansion in L, likewise included Harpocration in L'.20 This was, however, denied by Adler; 21 in order to explain why Harpocration glosses sometimes break the alphabetical sequence in Photius, and why they do not often appear in combination with other glosses, she supposed that Harpocration was used directly by the Suda and Photius (one must suppose also by B, though she does not say so). Erbse follows this without discussion.22 Alpers suggests a compromise: Harpocration may have been added to 1:.' but also used directly by the Suda and Photius.23 Theodoridis in his marginal notes to Photius implicitly follows Wentzel. To endeavour to settle this matter is the aim of this paper. It seems that the best chance of doing so is by a detailed examination of the glosses in a, where we have the benefit of three sources. 17 This is the traditional view, formulated by Reitzenstein and elaborated by Erbse. Alpers, BZ 64 (971) 80, argued that Erbse does not prove the the existence of 1:'; this may be so for the Atticistic glosses, with which Erbse is concerned, but cannot be maintained in general; what is needed is an explanation of the numerous glosses that appear in B Suda Phot., but not in A (C). IS Theodoridis (supra n.15) I Ixi-Ixxi. 19 GGA 155 (1893) 28; SitzBerlin 1895, 480. 20 Supra n.16: xxxiv-xxxix. 21 GGA 185 (1923) 28; RE (supra n.16) 692f. 22 Supra n.16: 24 n.3. 23 Supra n.16: 73 n.25. CUNNINGHAM, IAN C., Harpocration and the "Synagoge" [Greek] , Greek, Roman and Byzantine Studies, 27:2 (1986:Summer) p.205 208 HARPOCRATION AND THE I.YNAfflrH 2. Data: Distribution oj Glosses The glosses of Harp. appear in B Suda Phot. almost in their en­ tirety. For a there are 285 glosses in Dindorf's edition, but of these 14 are totally omitted in the epitome,24 and hence in B Suda Phot. (with two exceptions, a-yvEvETat and ap'}'Vp'iTt~ yi). Of the remaining 271 only 6 do not appear in any of them: clKtVaK71<; (18.12) (L' has another gloss, Bach. 54. 11 , Sudaa882, Phot.a754) clIL/JAw8pi&ov (25.13) (L has another gloss, Boysen Xla.18, Bach. 79.2, Sudaa1524, Phot.a116I) a~E1rw<; (29.10) (L has another gloss, Boysen XIIIb.21, Bach. 82.5, Sudaa1645, Phot.a1391) clvaTi (33.12) (L' has another gloss, Bach.85.19, Sudaa2102, Phot. a1691)25 aV-r}KEt (37.3) (L' has another gloss, Bach.96.25, Sudaa2405, Phot. a1917) aOtKo<; (38.10) It is noteworthy that in five of these cases a gloss from a different source is in I.; although in general two or more glosses on the same word are seldom avoided in Byzantine lexica, this does occasionally happen,26 and may be operative here. aOtKO~ is grossly out of place (presumably whoever first positioned it either read or was thinking of aVOtKo~), and from this an explanation of its omission may be found: it was to be restored to its proper alphabetical place but was over­ looked at that point. There remain 265. 162 of these are in all three of B Suda Phot. A total of 20 appear in only one of the sources, while 83 appear in only two. Before these are considered, it is necessary to note that in B the section from A,.L to AvS (middle) does not contain any addition from I.' or I."'.27 Glosses from that area in Suda and/or Phot. are there- 24 These are 34.13 av8pa'ITo&uTT/~, 40.16 'AVTu/lWV, 40.18 'AVTi'ITaTpO~, 44.10 aTTiA­ AEtV, 46.15 a'lTOtKia (omitted by D but not by E), 47.9 a'ITOA'T/gt~, 48.13 'A'lTOAAWV 'lTaT~~, 49.3 am> 1-£W'8wI-ulTwv, 56.4 apyvp'iTL~ yij, 56.7 'ApytvoVcrat, 56.12 'Apyaw~, 56.15 'Apwf3ap'&.V'T/'> (omitted by E according to Bekker, who does not give a full report of D; Dindorf is silent), 57.12 'ApUTTvUa, 68.18 'Al/xxPE-u... For a possible ex­ planation of some of these omissions see J. J. Keaney, GRBS 14 (1973) 418. 25 The addition in Phot., which resembles Harp., is most likely from Diogenianus; cf Hesychius a4636. 26 I am not aware of any general discussion of this, and I have not myself collected examples. One simple case that has come to my attention is this: the 1: gloss EI-''lTi~' KaJVwl/I (Bach.218.5) is not in Suda, unlike the vast majority of 1: glosses. Instead Suda E 1 020 has EI-''lTi~' KwvWl/It 'lTapa'ITAT,uwv from an Aristophanic scholium. 27 Reitzenstein (supra n.16) xxxiii. CUNNINGHAM, IAN C., Harpocration and the "Synagoge" [Greek] , Greek, Roman and Byzantine Studies, 27:2 (1986:Summer) p.205 IAN C. CUNNINGHAM 209 fore not significant~ this reduces the number to 12 in one source and 49 in two. B: 'A7TaTovpux (42.18; Bach.113.5) "Ap(JI-UO~ (57.10; Bach.143.3) a~ (67.6; Bach.162.25) Suda: a8vvaToL 01.12; Sudaa540; Bach. 32. 13, PhOLa393 from Bekk. v) aEi 02.6; Sudaa605; Bach.33.25, PhoLa405 from Phryn,) 'AII(JrW71 (38.4; Sudaa2513) allTt-ypacpi, {39.10; Sudaa2661)28 a~ollt (41.17; Sudaa2833)29 a7TOSEK'Tai (46.4; Sudaa3281; Bach. 124.2, Phot.a2480 from Bekk. v) a7ToO''TOAEL~ (51.5; Sudaa3559; Bach. 133. 16, Phot.a2660 from Bekk. v) apyvpLTL~ yiJ (56.4, not in epitome; Sudaa3792) a'Tif.L71'TO~ a-yoJll (64.1; Sudaa4364; Bach.160.3, Phot.a3094 from Bekk. v) Suda PhOL: af3tO~ (3.1; Sudaa47, PhOLa38; Bach.6.3, PhOLa37 from ?) a-ya(J7j~ 'TVX.71~ lIEW~ 0.6; Sudaa 111, PhoLa69) a-YllwJLIi, a'TTonJ.Wv, a'TTOTHxiO"at, (a'TToTil-'TlO"tV 1:), ,\ ,..., , a'TTO TOV 'TTpaYJ.UX.TOli Bach. 138.24-139. 1 a'TTpoO"KATlTOV, a'TTpol3ovAEvTov, a'TTpoO"TaO"wv, a11'pOTWV Bach.141.23-142.6 apyvpis 8-r1K"fJ, apyvpoK011'E'ioll, apyvpo11'OOO Bi­ ¢pov, (apyvpiov BiK"fJ Bekk. v), "ApyovO"a Bach.161.17-22 aTTa, "ATTTlli, 'ATTtKOLIi ypal-'J.UX.O"tV and one long one: Bach.103.24-105.16 'Avnoxi~, (allTtXOpBali 1:'), "AVTPWVE~, (av­ TW/.LOO"ia Tim.), aVTW/.LOO"ia, (avTW/.LOO"ia 1:'), 'AvTu!tTa~, avn­ I3A'YJ8EVTa~, 'AlITtYE"ioo~, allTtypa¢E~, allnypa~, (avTi8ETOV Phryn.), avn8EO"tli, (avTi8EToV ?), 'AvnKVpa, "AvnO"O"a. This last comes between the 1: (some 1:') aVT- series and the 1:/11 one. Apart from these most of the Harp. glosses are not grouped in any significant way. There are, however, a handful that appear in dis­ tinctly r"' contexts: ayvEvETat (Bach.24.1) in its correct alphabetic place in the long (about 75 glosses with scarcely an interruption) 1:'11 ay- series 'AptO"TEV<; (Bach.143.8) in a small group of four glosses apPTlpELV (Bach.l45.24) near the beginning of nine app- glosses mostly from 1: III 'ApxtMJJ.EWIi 'TTOAE/.LOIi (Bach.l49.18) in a run of about the same length aO"Ta81-'TlTOTaTOv, aO"nKTov xwpiov, aO"TVvO/.LOt (Bach.155.16, 19, 23) near the beginning of a group of a dozen. Of these ayvEvETaL is in Phot. but not in Suda or in the epitome of Harp.; 'ApLUTE~ is in Phot. but not in Suda; appTJcf>opE'iv and 'APXL­ sa~ELO~ 11'OAEJ..LO~ are in Suda but not in Phot.; and the other three are in both Suda and Phot. 30 Reitzenstein (supra n.16) xxxiii-xlii. CUNNINGHAM, IAN C., Harpocration and the "Synagoge" [Greek] , Greek, Roman and Byzantine Studies, 27:2 (1986:Summer) p.205 212 HARPOCRATION AND THE LYNArnrH Order in Photo The breaking of alphabetical order by Harp. glosses in Photo was, as already noted, one of Adler's reasons for thinking that they were added directly by Photo Reitzenstein remarked briefly on the alpha­ betical order that "kein VersHindiger wird darauf verzichten, aus der Stellung bei ihm ab und an auf den Ursprung der Glossen Schllisse zu machen-immerhin aber so, dass eine nur von ihm ausgehende Untersuchung nie zu klaren Ergebnissen kommen kann."31 The anal­ ysis begun by him (xlii-xlv) can now be extended. While Photo is much closer to being fully alphabetized than for example B, it is (in contrast with Suda) far from completely so. Of the two-letter sequences in a, only the brief aa- (17 glosses) and a'YI- (11) have no irregularities. A very common reason for minor breaches of order is that inflectional endings and the like in words from the same root are not taken into account; particularly extensive examples are c:lyavaK'T- 100-03, atp- 639-53 (649 inserted), c:lKOV- 802-21, c:lJ.UlX- 1149-57, and c:lv(JpW'Tf- 1974-89. Unamalgamated se­ quences can be seen in c:l8w- 346f/348-53/354/355-59; c:let- 404-16/ 420-23 (c/. below); aip- aln- ai'T- 637-73/686-98; c:l~V- 1258-84/ 1285/1286-90; c:l~- 2172-76/2179-82; c:l1TOA- 2535-49/2550-59. Some out-of-place runs can be recognised as being from L''': 102-17, 147- 51, 211-15, 355-59, 471-75, 505-19, 617-34, 755-58, 822-30, and 3313-18. Single glosses or small groups that are distinctly out of place are 93-97, 486, 558, 615, 649, 654, 655f, 660, 684, 685, 799f, 917, 982, 984-86, 1039-42, 1136-43, 1225-32, 1347~ 1471, 1644-47, 1683-85, 1798, 1800-02, 1941, 1973, 2051-66, 2112f, 2125-29, 2135f, 2162, 2276, 2296, 2298, 2301, 2303-07, 2365~ 2521, 2527, 2561, 2650-58, 2659-65, 2708, 2745f, 2775-77, 2780-82, 2866-69, 2871f, 2873f, 299Of, 3210-12, 3216-19, and 3248. There are many smaller irregularities. Of the Harp. glosses that disturb the alphabetical order many come into the category of words from the same root: 73 'Aya(}apxo~, 69 eXya6ii~ nJX71~ IIEWt;, 226 eXyopa~, 230 eXyo­ paUat, 260 Ctypiov<;, 279 CtYVw~, 318 CtywIIWII, -WIl7"E~, 521 Al­ yEW II , 522 AlYEtOOL, 534 ai8Euau(JaL, 587 aiK£aS, 716 CtKap, 1090 'AAw'TrEK71(}EII, 1091 'AAW7TEKOIlIl71UO~, 1400 eXlla!J&AAEw, 1472 eXlla- 8Eu(}aL (1473 is a later insertion), 1493 eXllaillEu8aL, 2089 eXlI'TLypa­ cf>EV<;, 2422 eXm)XEta, 2570 eXTrO/JMT(}OL, 2608 eXTroTroJ.'Tra'i'ot, 2790 eXPYVPOKOTrEW II , 2815 'ApW7"EV<;, 2825 eXpK7"EVUat (or 2824 may be out of place), 2929 eXPX~ all8pa 8EiKIlVUt, 2953 eXU711UXl17"a, 3094 31 Supra n.16: xxxi. CUNNINGHAM, IAN C., Harpocration and the "Synagoge" [Greek] , Greek, Roman and Byzantine Studies, 27:2 (1986:Summer) p.205 IAN C. CUNNINGHAM aTilJ-'TITO~ aywv, 3136 'ATTLKOt~ ypalJ-f..UX(ItV, 3176 aVAEw~, 3378 O:CPOPIJ:rl, and 3428 O:xapu:rrELv. 213 Others, despite initial appearances, do not in reality disturb the order: 223 a-YOL (221f have been added), 269 'AypOLAijOU (267f are out of place), 1098f 'Ap,af>OKO<;, 'AJ,£a'ovwv (1097 is out of place), 2082 'AVTLYEvi8a<; (2083-85 are out of place), 2196 a07T'Ta (2195 is out of place), 2350 Cx7rEpyafTeXf.l,Evo<; (2349 is out of place), 2675 Cx7ro'Ta~,<; (2674 is out of place) and 3234 aVTOIUXXELV (3233 is out of place). The following do disturb the order: 130 'AyaULKAr,<;, 285 'A"vppw<;, 354 a8wfTTa'Tov (e! above), 374 aooKiJ.UX(rTo~, 422f aELECrTW, aELAoyia (with 420f a separate r." block), 468 'AOJ..LOVEV<; (467 is the previous Harp. gloss), 605 Al­ vio~, 744 "AK..,." 893 'AAaLEV<;, 915 'AAea~ (916 is the next Harp. gloss), 984-86 'AAKETat;, 'AAKLf3w8'T1~' 'AAKif..UXXo~, 1039-42 aAoyoL EpavtfTTai, o:i\oyiav, 'AAoVV1]UOt;, 'AA07r1J, 1171 'A,.L{3paKia, 1213 a,.u/'l""OL (e! 1212 alJ-L'Tf'7TOv) , 1316 alJ4>t8eat; (1317 is the next Harp. gloss), 1500 'AvaKaiaUtV (1499 may be a later addition), 16100:va7ro8"oJ,LEva, 1879 o:vEuKEvauavTO, 1946 "AvOEta (1947 is the next Harp. gloss), 2135f 'AVTWXIS, "Av'TLuua, 2150 aVTwJ..LOuia, 2484 a7T08L&)J,LEVOL, 2631 a7TOUa~aVTa, 2660f a7TouTOAEtt;, a7TOU'T1J­ uaJ,LEvov (in a confused area), 2687 a7TOTELxiuaL (in a confused area), 2703 a7ToqxxUtt;, 2715 O:7Tocpopav (in a confused area),32 2746 O:7TPO{3oVi\EV'TOV (e! 2745 O:7TPO{3oVi\ia), 2795 "Ap81]'T'Tot (the Bekk. v gloss on the same word is correctly placed), 3130 "AT'TL~ (in a confused area), and 3322 acp' 'EUTiat;. That is, a maximum of 38 glosses out of 227 are out of place. Given the character of the lexicon that does not appear an excessive number. Combined glosses The manner in which similar glosses of different OrIgm are com­ bined is a further means of establishing relationship. The Harp. glos­ ses concerned are these: O:YEi\a'iot;/aYEi\aiwv. There are four glosses: (1) O:YEi\aLOt;· i8~<;. Tj <> EV ayeATl 8wywv (I.: Cyr.+AeI.Dionys.aI7), (2) aYEAaiwv' l8LWTWV ... (r.': Cyr.+?), (3) aYEAaiwv' 'TWV 7TollWv Kat roxov­ TWV . .. (Harp.4.l 0), (4) aYEi\atOt;' aVTi TOV i8uV'T1Jt;, pEJ,L{3W81]t;, Kat EV'TEi\-r,<; ... (I."': Paus.a12). These are combined as follows: B (Bach.8.25f) (2)/(1)+(3)+(4); Sudaa187 (1)+(2) (+Julian) + (3); Phot.a134, 141 (2) + (3)/(4). 32 2716f appear to be placed according to the antistoechic order used in Suda, of which there are traces elsewhere; cf Adler (supra n.16) 679. CUNNINGHAM, IAN C., Harpocration and the "Synagoge" [Greek] , Greek, Roman and Byzantine Studies, 27:2 (1986:Summer) p.205 214 HARPOCRATION AND THE I,YNAfOrH ayopavo~c.. B (Bach.14.7) and Sudaa302 have the Harp. (6.8) gloss by itself, Phot.a229 adds a sentence from Bekk. v. Alvwv<;. The epitome of Harp. (17.9), by omitting the citation from Demosthenes (23.119) that followed the lemma, produced the rather odd-looking gloss Alvwv<;' AlvOt; 1TOAt~ EUTt 'T'ijt; 8p~­ K'YI~ .... This is reproduced by B (Bach.47.4) and Phot.a605. Suda misunderstood it and (combining it with another source) produced (at224) AivtOv Kat Alvwv<;' m>AEt~, (at225) Alvo~' , :t''''' a ' 1TOAt~ EUTt 'TTIt; 'C1pl!tK'YI~ .... aKpoau8at. Sudaa993 , Phot.a849 have from Harp. (19.11) aVTt TOU lnraKOl)Etv. oVr~ 'AVT~V (fr.62 Th.). For OVT~ B (Bach. 56.12) substitutes Kat lnrOTET(!t-X.8at· TIAthwv ( Grg. 488c) Kat from Phrynichus (Praep.Soph. 38.6). aAcluTwp, There are three glosses: (1) q,oVtKO~ OOt/UUv, Tt/UUpwv Kat aVEA1TtUTa 1TOtWV (I.), (2) <> TOtaWa TOA/LT,Ua~ 6Jv /LT, EUTtV E1TtAa8EU8at (Harp. 20. 9) , (3) <> alJ.apTWAOt;, KaTtl ""EV XPVCTt1T- 1TOV ••• (Paus.a61). Phot. (a903, 902, 896) has them separate. B (Bach.65.7) inserts part of (1) and all of (2) near the beginning and end respectively of (3). Sudaa1082 has (1) with 1TtKPO~ Kai inserted from & + extract from Polybius+ (2).33 &.At1TEOOV. Two glosses (I. and Harp.21.18) in Phot.a955f are com­ bined by insertion of 8E in B (Bach.67.5) and SudaaI240. av8pa1TOOOKcl:'7T'rJAOt;. Two glosses (I. and Harp.34.1O) are combined by insertion of Kat in Phot.a1746, of o~v (with an Aristophanic scholium and an excerpt of Procopius) in Sudaa2155. 'Av8EU'T"fIpUuV. In the gloss from Harp. (38.1), Sudaa2500, Phot. a1955, there is one etymology of the word: B (Bach.97.23) inserts an alternative. a1TayW'Yll. Two glosses (I. and Harp.42.8) are grouped in one of the Suda's characteristic paragraphs, preceded by the I. gloss a1TayW'Ycl:~ and separated by an excerpt from Philostorgius (Su­ daa2869). In Phot.a2208f they are separate, but the former has an addition AEYETat a1TayW'Y'l1 Kat T, aiXJ.UlAwuI.a Kat ii TtVOt; KOIJ.tM7, otov T, aywyr). In B (Bach. 109.20, 24) this is added to the Harp. gloss, whose last sentence is transferred to the other. a1TapTtAoyl.a. Two similar glosses in B (Bach. 11 1.23 d a'7T'rJPTtU~­ VOt; apt8~t; Kat AOyot;. cHpOOOTOt; ifJ&}J,Up [7.29.2]' AEyEI. 8E 2EpfrJ~ ... ; 111.29 aVTi TOU a'7T'rJPTtU~vOt; Kai 1TAT,P"'lt; apt8- ~t;. oVr~ AVUl.at; [p.334 Th.1 Kai cHpoooTO~), the latter from Harp. (42.16), are combined in Sudaa2929 a'7T'rJPTtU~vOt; Kai , , 8' , " ' ., A ' cH' ~- ~, y' " ' 1TA"'IP"'l~ apt ~~ Kat AOyO~. OVT~ VUtat;. POouTOt; oE ':. . AEyEt 8E 2EP~t; .... 331 need not consider the vexed question of Etym.Gen.; see Alpers (supra n.16) 76; Theodoridis (supra n.15) xl-xliv. CUNNINGHAM, IAN C., Harpocration and the "Synagoge" [Greek] , Greek, Roman and Byzantine Studies, 27:2 (1986:Summer) p.205 IAN C. CUNNINGHAM a1TEf,1TE'V. Phot.a2311 has the Harp. (43.7) gloss aVTI. TOV a1To­ Ka#LE'V Kat aSvva'T"lluac.. Eun BE Kal. aVTt TOV a1TapvTwau(Jaf,. Sudaa3124 has it with ij a1TEf,1TE'V' substituted for Eun SE Kat aVTt TOV. In B (Bach.116.7) it has the same form as in Phot., but 7j Kat a1TtUTEtv is inserted after aBvva~uaf, (by confusion with the gloss Sudaa3116, PhoLa2311 a1TE/,(JEtv' a1TwTELv). (17rEaXOUlW~1I0'). B (Bach.118.26) has the 1: gloss Cc11'f.(TX.OWtcTJd­ vov· O:1T£UTEPT/p,EVOV, 0:1TOK£KAEWp,EVOV. The epitome of Harp. (44.3) has a1TEUXOf,VWJ-tEVo~' ~T/f.WU(JEVT/~ EV TeP KaT' 'Apf,UTO­ YEt.TOVO~ (25.28) aVTI. TOV a1TOKEKA.EWp,EVO~. PhoLa2375 seems to be this, with the citation omitted. Sudaa3079 combines the t ' " '... (J' , " wo: a1TEUX0f,Vf,U#LEVO~' a1TEUTEPT/#LEVO~. aT/f.WU EVT/~ EV TCY Ka- Tel 'ApWTOYEiTOVO~ aVTi TOV a1TOKEKA.EWJ-tEVO~. a1Toypaqn.,. B (Bach.l22.25) prefixes the ~ (Cyr.) gloss to that of Harp. (45.1 O. Sudaa3273 has them in reverse order. Phot.a2468 has only the former. a1To1TEq,aUJJ.ElIOll. B (Bach.130.l4) and Sudaa3475 join the ~ (Cyr.) and Harp. (49.11) glosses with ii. Phot.a2604 has only the latter. a1TouToAEL~. Sudaa3559 has the Harp. (51.5) gloss oi E1Tt ~~ EK- 1TOfJ:rrij~ TWlI TPL'J1PWlI a1T08E8Ef,yJJ.ElIOL. PhOLa2660 has the Bekk. v (203.22) gloss f,' TOV apf,(Jp,av -ryuav apxollTE~ E1Tt TWV EK1TAEOV­ UWV TPf,T,PWlI Kat TWV allayoJJ.ElIwv UT()Awv. a1TouToAE'~ BE 1Tapa TO a1TOUTEAAEf,V Ta 1TAOUx. B (Bach.133.16) also follows Bekk. v in essence, but inserts material from Harp. with oi E1Tt, rii~ EK- 1TOfJ:1TTi~ TWV 1TAEOVUWlI TPLT,PWV Kat, TWlI avayoJJ.ElIwv UTOAwll a1TOSESEf,yJJ.ElIOf,. apya~. B (Bach.141.14) and Sudaa3760 have the Harp. (54.6) gloss at the end: ~WPLE'S Kat pAiAWTa 'ApYE'Of, Tall Oc/xll apyall EKa­ AOVV .... Before this B has 0 SEf,VOTaTO~ 1Tap' r,Af,Kiav apya~ KaAEtTac.. OVT~ lAeyov Kat, TOV ~T/f.WUBEVT/V. aAAOf, BE q,aUL TOV­ TOll KaKOll 1TOLT/T1JV YEyovellaf, (then the inept linking phrase oi BE OTt KaTa). Suda has <> B. 1Tap' r,. a. K. TOVT~ acpof..WtOvUBaL inrO Aluxillov TOV ~T/f.WuBEvT/v. aAAOf, BE TOV BEf,VOTaTOll Kat BT/puUST/ TOV TP01TOV (omitting ~Wpf,E'~). B has clearly incorpo­ rated material from the Bekk. v (206.7) gloss in PhoLa2768, oi J-tEV cpaUf, TOVTOV KaKoll 1TOf,T/~lI YEyollElIaL, oi BE Tall ap­ yav ocJ>e~ ElM~ cpaUf,. The origin of the main insertion is not known.34 apf.WUTai. As with a1TouToAE'~, Sudaa3979 has the Harp. (58.16) gloss (oi 1m0 TWV AaKEOOf,f.Wviwll el~ Ta~ lnrTjKOO~ 1rOAE/,~ ap­ XOllTE~ EK1TEJL1TO#LEVOf,), Phot.a2838 the Bekk. v (206.16) gloss (oi cppovpapxof, TWlI AaKEOOLf..WlIiwv, 1Tapel 'TO apJLD'ELlI Kat Ka­ BWTav Ta~ 1m' airrwv cpvAaTTOJ-tEVa~ 1TOAE/,~), and B (Bach. 215 34 Two MSS. of the full Harp. have the same as Suda, but omitting TOVTCP ... ll.7If..W­ UfJEVrt" at the end of the gloss: interpolated from Suda? CUNNINGHAM, IAN C., Harpocration and the "Synagoge" [Greek] , Greek, Roman and Byzantine Studies, 27:2 (1986:Summer) p.205 216 HARPOCRA TION AND THE I.YNAfilrH 14S.2) combines the two (oi lnrO TWV AaKEOOLf.£Oviwv El() T(Z() It , '\." 'A..' '1' , 1I7T7IKOOlJ() 1TOn.EL() apXOVTE() Kat o/povpapXOL EK1TEP.1TO!-£EVOL, 1Tapa TO ap,.,.o'ELV KTA.). apxaw(). Here two glosses, apxaw()· aVTi TOV Eirr,fh,() IIMhwv (Hp.Mi. 371D~ perhaps from Phrynichus fr.263), and apxa~· 'IuoKpaT71() ,.uv ... (from Harp.60.S), are differently treated. Phot.a2920f has them separate; Sudaa4074f has them separate but in reverse order, with the latter preceded by the Ambrosian <> 1Ta'l\.auk B (Bach.l48.14) runs the two together. It is to be noted that all three have, explicitly or implicitly, the lemma apxaw() in Harp., for the original apxa~. aUTpa/3TJ. Phot.a3017 has the gloss TO E1Ti TWV Ecpi'1T7TWV ~AOV, 0 ~ '8 y' ,,\, ~ ~ r' t' KpaTOVUtV OL Ka E.,,o!-£EVOL. KaL aVTo TO VWT0'fIVPOV V1TO."VYWV. Ot 8E TOV UWp.aT71,,/Ov TJ,."wvov aUTpa/3"f1V EKaAEuav. Sudaa4248 runs together the L gloss OVO(), the Harp. (62.18) gloss TJ TJ,."w­ VO(), ~ 4.TJf.£OU8EV'T1() KaTlZ MELSwV (21.133). Taxa 8E Kai 1Tav 'y' ',,-' ~ ," 8 ' ~ "'" ~ d V1TO."vyWV, Eo/ ov Ot av PW1TOt oXOVVTat, OVTW() EKan.EtTO, an the gloss found in Phot., interposing between the two sentences of the latter aUTpa/3"f1AaT71()· 0 TJVWXO() from the Ambrosian lexicon. B (Bach.1S4.14) joins the gloss in Phot. with Harp.: TO , \ ,.. 'A..: l:,.,.'.\. f' r' \,.. I' r' ',1..'''' Em TWV E'I"''1T7TWV svn.OV ... V1TO."VYWV. Kat 1Tav V1TO."V,,/WV Eo/ OV oi av8pW1TOt WXOVVTO oWW() EKaAE'iTo, ~ Kat 4.TJf.£OU8EVTJ() KaTlZ MEt8wv. Abbreviated glosses A few glosses appear in one source or another in an abbreviated form. In B: "A{3apt(). Harp.1.l, Sudaa 18, Phot.a29 have a long narration of his history. B (Bach.S.1S) has only the opening words ovop.a ,ropwv. 'A/3POKOp.a(). B (Bach.S.2) again has only ovop.a ,ropwv. Harp. epit. and Phot.a54 add uaTpa1TTJ~ 8' ,Jv 'ApTa~Ep~ov TOl) IIEpuwv {3aULAEW(). The full Harp.3.3 and Sudaa83 have otrO() uaTpa1TTJ() ,Jv KTA. 'A,,/a8apxo~. The final phrase TO 8E ,,/EVO() Lap.to~ in Harp.3.4, Sudaa109, Phot.a73 is omitted by B (Bach.6.20). In Phot.: aVE1T01TTEVTOV. Phot.a 1862 omits the citations from Philochorus (FGrHist 328F69f) that are given by Harp.36. 7, Sudaa2303, and B (Bach.91.11). aVTL,,/pat/JEVc;. Harp.39.S, Sudaa2661, and B (Bach.104.2I) have <> Ka(}LU'Tal-UvO~ E1Tt TWV KaTa{3aAAovTwv TLVlZ rii 1TOAEt xpr,p.aTa, " , '~(} " " ... 8' , ~ , ''A WUTE aVTL,,/pa~u at aVTa. OVTW aTJf.£OU EVTJ~ EV T~ KaTa v- 8pOTiwvo~ (22.70). aAAa Kat A1nxi.v"f1~ EV Tcii KaTlZ KT71UuPWVTO~ (3.25). 800 8E ,Juav, <> ~v rij~ 8WtK7/UEW(), <> 8E rij~ f30vA-ij~. CUNNINGHAM, IAN C., Harpocration and the "Synagoge" [Greek] , Greek, Roman and Byzantine Studies, 27:2 (1986:Summer) p.205 IAN C. CUNNINGHAM PhoLa2089 has only the last sentence, with the addition ~ cfn1UL 1l.T/,.wCT(JE,,1T/<; Kai AiCTXivT/<;. a7TEuXOL VLUJ,U 1.10<;. See above (215). a7Ti\a<;. Phot.a2435 omits mention of Callistratus from Harp.45.1, Sudaa3223, and B (Bach. 121.22) . apE'T-q. Harp.57.9, Sudaa3831, and B (Bach.142.21) have 'Av80- Ki8Tf<; (fr.6) Kai @ovKV8i8Tf<; 0.33.2) aV'Ti 'TOV Ev8o~l.a. Phot. a2802 alters this to ~v8o~l.a. o{;'Tw<; 'Av8oKl8T/<; Kai @oVKV8l8T/<;. In Suda Phot.: a7TaYE. Harp.42.8 and B (Bach.l09.14) have aV'Tt 'TOV )(pw rfi a7Tayw­ yfi. oiiTw<; dTfJ..WU{JEVTf<; EV 'T~ Ka'T£l 'Av8po'TL(J)J.IOfO (22.26). {UTL 8E 8iKTffO Eli)ofO. Sudaa2861, PhoLa2201 omit all after dTfJ..WU{JEVTffO. Textual variations 217 The relation of the r. sources to the epitome of Harp., plain from the selection and wording of the glosses, is confirmed by the textual variations. B Suda Phot. Harp.epit.: Harp.plen. ayvw<; (Harp.7.8, Bach. 16.8, Sudaa383, Phot.a279) CT'Tai\ay,.wv<;, {3ap{3Ctpovc;: CTTai\a'YJ..WL<;, /3ap{3Ctpwv 'A8paU'TELa (Harp.lO.16, Bach.28.15, Sudaa524, Phot.a385) TaV- , , " ,,....', 'TOV: TTfV aV'TT/v; aV7f/. aVTTfv aEpK'TOV (Harp.14.1, Bach.34.l 0, Sudaa560, Phot.a424) 0PW/LEvov: Ka{J0pW/LE 1.10 v aiKI.a<; (Harp.17.5, Bach.44.9, Sudaad69, Phot.a587) 7Ti\T/'YaL<;: 7Ti\Tf- " YTf<; AivioVfO (Harp.l7.9, Bach.47.4, SudaaL225, Phot.a605) E7TTfyayoV'To post KV/L T/fO: post ML 'TVi\ -q VT/<; ~wv71ULl.l (Harp.18.1, Bach.47.7, SudaaL242, Phot.a609) i\EYECT­ (JaL: i\EYELV aKap (Harp.18.11, Bach.53.1O, Sudaa800, Phot.a716) aKap: aKap71 vel aKapEi 'AK'T-q (Harp.19.12, Bach.60.4, Sudaa1036, Phot.a876) i8iw<;: om.; 7Ti\Eiw: 7Ti\Eiw J..WLpav "AK'TLa (Harp.20.2, Bach. 64. 11 , Sudaa1037, Phot.a877) 'T~ 7TEpi ,... , 1"'\ , 'TWV: EV 'TqJ 7TEpt 'AAE~av8pofO (Harp.21.3, Bach.66.6, Phot.a916) Kat: Ei There are, however, a handful of cases where one or more of them agrees with the full version against the other(s) and the epitome. Photo Harp.epit.: B Suda Harp.plen. AiYEL8aL (Harp.15.11, Bach.42.1O, Sudaad2, Phot.a522) YV-qULOfO ECTTLV: 'YV-qCTW<; CUNNINGHAM, IAN C., Harpocration and the "Synagoge" [Greek] , Greek, Roman and Byzantine Studies, 27:2 (1986:Summer) p.205 218 HARPOCRATION AND THE ~YNAfnrH Suda Phot. Harp.epit.: B Harp.plen. aVTL'Ypac/n1 (Harp.39.1O, Bach.I04.26, Sudaa2661, Phot.a2090) lem- hb " , ~, , 1""1 , rna om.: a.; Ta ante EV: ante Tell"; npa'YI.UXTO<;: '1TEpl. TOV '1Tpa'Y- /UlTO<;35 'ApKTOVP0<; (Harp.58.10, Bach. 144.20, Sudaa3961, Phot.a2827) BE Kat: 8E 'Ava<; (Harp.70.5, Bach.l73.25, Sudaa4659, Phot.a3406) 6vo/Ul­ Ta: OVO/Ul Suda Harp.epit.: B Phot. Harp.plen. aUTUCTov XwpWV (Harp.62.14, Bach.155.19, Sudaa4226, PhoLa3015) aUT . . . : aUTLKTO v There are also cases where they all agree against Harp. (generally both versions): B Suda Phot.: Harp. 'AyYJufAao<; (Harp.5.3, Bach.12.25, Sudaa229, Phot.aI7I) BE: BE \ Kal. a'Y01. (Harp.4.3, Bach.13.23, Sudaa381, PhoLa223) a'Y0I.: a'Y0L TOV­ TO aVEAoVua (Harp.35.9, Bach.90.1, Sudaa2248, PhoLa1836) aVrr1v: alnfi 'AVTL'YEvi8a<; (Harp.39.3, Bach.I04.l9, PhoLa2082) dLOVVuOV: dLO-, VVULOV aVTL'YpaEv<> (Harp.39.5, Bach. 104.21, Sudaa2661, Phot.a2089) ... ... , ,,1.,.,"" 'YIuav: 'YIuav aVTL'Ypa~L<; aVTwJ..£Ouia (Harp.41.4, Bach.l04.1, Sudaa2759, Phot.a2150) 'Ypa­ l/JavTwv: 'Ypal/JavTE<; avwp8pia,ov (Harp.41.8. Bach.107.3, Sudaa2598, Phot.a2167) av­ wp8pia,ov: avwp8ia,ov; Op8PLOL: Op8LOL (epit.: Op8LOV plen.) a'1TpouTaUWV (Harp.53.12, Bach.138.28, Sudaa3703) -r,PVELTO: !lPEL­ TO aP'YVpoKO'1TELoV (Harp.56.5, Bach.141.27, Sudaa3796, PhoLa2790) " , 'YI J.Ul "TTl pLO V: UTI J.Ul V TTl PLO v apKTEVuaL (Harp.58.4, Bach.143.23, Sudaa3959, Phot.a2825) '1Tap- 8 ' 8'" E VOl.: '1Tap E VOl. apKTOI. apP'YItPopELV (Harp.59.l, Bach. 145.23, Sudaa384S) apP'YIVotPoPELV: apP'YItPopEL v apxaw<; (Harp. 60.5, Bach. 148. 14, Sudaa4074, PhoLa292I) apxaL- 0<;: apxaiw<; 'ApXI.Mp,ELO<; '1TOAEJ..£O<; (Harp.61.6, Bach. 149.18, Sudaa41OS) TOV OEAO'1TOVV'YIUUXKOV: om. epit. (hab. plen.) 35 B perhaps took these readings from the related gloss Bekk. v 200.9 (Bekk. v and Harp. share a source in the so-called Onomastikon: Wentzel, GGA 159 (1897) 618). CUNNINGHAM, IAN C., Harpocration and the "Synagoge" [Greek] , Greek, Roman and Byzantine Studies, 27:2 (1986:Summer) p.205 IAN C. CUNNINGHAM acPEt,1) Kat, a1Ta?t.?t.a~al) (Harp.69.I, Bach. 170.3, Sudaa4599, Phot. a33I9f) Ttl): Ttl) Ttva 219 Errors peculiar to only one of the 1:. sources are obviously of no significance here (even less, variations within Suda or Phot.). But there are some that occur in two of them. Suda Phot.: B Harp. (prima facie error in 1:.") 'A8J.f,OVEV<; (Harp.lS.7, Bach.37.16, Sudaa743, Phot.a468) 'A8IUJJv-: 'A(JJ.f,Ov- "AxY, (Harp. 19.1, Bach.S3.29, Sudaa8S8, Phot.a744) KaAOVJ,£EVTJV: vvv KaAOVJ,£EVTJV amJcpauIS (Harp.52.I5, Bach. 136.24, Sudaa3629, Phot.a2703) Ka- \ ,., , Av.!.. • \'" , Av.!.. ~EtTat ... a1TO'Ypa'f"" Ka~Et ... a1TO'Ypa'f"'V "AT7T/I) (Harp.65.4, Bach. 161.20, Sudaa4354, PhoLa3130) "ATTtI): " AT'TT}I) acpopl-'T, (Harp.69.1S, Bach.l72.30, Sudaa4638, Phot.a3378) SWuEL: 8Wun (~ Harp.plen.) B Phot.: Suda Harp. (prima facie error in 1:."') 'A'TJVtEV<; (Harp.14.6, Bach.35.I9, Sudaa594, PhoLa436) 'A{TJVEV<;: 'A{TJVtEV<; Aivl.ovl) (Harp.17.9, Bach.47.4, SudaaL225, PhoLa605) KaTCfJKTJUav: , KaTWKLuav a1TEp'Y~Ual-'Evol) (Harp.44.l, Bach. 117 .19, Sudaa3036, Phot.a2350) -: EV T4} 1TPOI) 'A1TOUOWpov a1TTJ?t.OTJI-'EVOI) (Harp.44.6, Bach. 12l.3, Sudaa31S6, PhoLa2408)-: , ,.., ' ..... ' EV T~ 1TpOI) ... aLVL1T1TOV ap'YVpil) fJT,KTJ (Harp.55.12, Bach.l41.23, Sudaa3792, PhoLa2786) 'R,!. a,l. Ta KL,..,...,TLa: Kt,..,...,TLa aVTOAT,KV(JOL (Harp.67.7, Bach.166A, Sudaa4505, PhoLa3227) hot­ J.f,OV!): ETO{,~36 B Suda: Phot. Harp. a'YEAatOI) (HarpA.I0, Bach.8.26, Sudaal87, PhoLa134) ptf3aTJv: pv­ aTJv a1To?t.E?t.oL1TOTEI) (Harp.47.6, Bach. 127.19, Sudaa3383, PhoLa2543) I.wKpa'TT}l): 'IuoKpaTTJI) 3. Conclusions It should be immediately apparent from the above data that simple solutions will not suffice. If Harpocration had been incorporated in 1:. 36 This is not quite the same as the other cases: the correct ETOil-Wv<; is also in Harp. pien., so that the error is due to Harp.epit. and was taken over by 1;' but corrected by l:'" . CUNNINGHAM, IAN C., Harpocration and the "Synagoge" [Greek] , Greek, Roman and Byzantine Studies, 27:2 (1986:Summer) p.205 220 HARPOCRATION AND THE l:YNArfirH at only one point, which would have to be l:', several difficulties arise. The considerable number of glosses in Suda Phot. but not in B has to be explained; B may abbreviate on occasion and combine freely, but cannot be shown to have omitted on any scale. Phot. is different in this respect (supra 207 and n.17), so that glosses in B Suda but not Phot. can be readily explained by omission in the latter. Further, the few glosses in B Phot. but not Suda-especially those in l:'" contexts in B-call for explanation. Next there is the lack of common combinations noted by Adler: only the simplest are found, in aAL7T€OOV and a11'07T€¢a(T~VOv. Finally, there are the textual al­ terations in B Suda: 'I(ToKpa1'T}~ s. v. a7TOA€AOL7T01'€~ might easily be a conjecture by Photius, but that in a'Y€AaLO~ is more recalcitrant. But converse problems are raised by Adler's proposal of indepen­ dent introduction to the sources. There are considerable overlaps in the glosses included by each, and the errors shared by two of the three are an almost insuperable objection Ot will be recalled that Adler fails to mention that Harpocration is included in B). Alpers' suggested compromise of incorporation by l:', Suda, and Phot. inde­ pendently does not fully meet the facts either: the glosses found only in B or B Phot. (with a l:'" context in B) and the errors common to B Phot. remain unexplained. There is no conclusive evidence that Bused Harpocration.37 Given that Phot. is extant in an abbreviated state, the three glosses found only in B may have been in l:'" and omitted in Photo The larger number in B Suda might be explained by independent use of Harp. by each, but in view of the conjunctive errors in B Suda it is easier again to suppose them to have been in l:' and omitted in Phot.; or to have been in l:"', omitted in Phot., and added independently by Suda. The handful of cases where B (sometimes with Suda) has a reading also in the full version of Harpocration, while Harp.epit. has another, may also be otherwise explained: S. vv. Ai'Y€L8aL, 'ApK1'OVPO~, and 'A~va~ the variations are trivial and may be coincidental, while S. V. all1't'Ypaqrq B may be contaminated with Bekk. V. Equally it is not clear that Phot. used Harpocration: Phot. by itself has no glosses outside the aJ.L- and av- sections where I:'" is not repre­ sented also by B. The reading pV8T}1I S. V. a'Y€AaLO~ is the best evidence. For Suda the case is much stronger: nine glosses not also in B or Phot., and one of these, apYUPL1't~ yij, not in the epitome of Harp. 37 There is no distinction to be drawn here between B itself and any intermediary there may have been between it and I."'. The same applies, mutatis mutandis, to what is said below about Phot. and Suda. CUNNINGHAM, IAN C., Harpocration and the "Synagoge" [Greek] , Greek, Roman and Byzantine Studies, 27:2 (1986:Summer) p.205 IAN C. CUNNINGHAM 221 The glosses that appear in B in I.'" contexts and are also in Suda are probably to be regarded likewise as separate additions to. the latter. Textual agreements with the full version of Harp. are, as already noted, trivial. When Suda agrees with Harpocration, and B Phot. have a variant, the possibility of Suda having by direct use of Har­ pocration corrected an error of I.' cannot be excluded; but error by r'" is at least equally likely. Use of Harpocration by r.'" is virtually certain from the occurrence of ayvEvETat only in B Phot. and decisively in a r.'" run in B, along with several conjunctive readings. For I." there can be no doubt at all: 22 unique glosses and several conjunctive readings. Equally so for I.', with the great majority of Harp. glosses common to all sources, and again with conjunctive readings. It appears therefore that Harp.epit. was used by r.', which incorpo­ rated most of its glosses. Many of those omitted (for whatever rea­ son) were added by I.", some others by r.'''. Consultation by Suda (and of a different version) is probable, by Band Phot. no more than possible. This complicated picture of the repeated use of the same source at several stages in the development of I. cannot be regarded as inherently unlikely. It is similar to the process Erbse has depicted for the Atticistic lexica. The historical circumstance to be imagined probably involved several scholars producing their own version of this collection of useful words, all working at roughly the same period in Constantinople and using the same limited number of sources. NATIONAL LIBRARY OF SCOTLAND, EDINBURGH June, 1986