Published online by Cambridge University Press: 09 November 2011
The Itinerarium Provinciarum Antonini Augusti is a collection of some 225 routes along the roads of the Roman Empire. In each case the beginning and the end of the route are given, together with the total mileage, and there then follows a list of the stopping-places on the route with a statement of the number of miles between each. The Itinerary begins in Mauretania, at Tingis, and then goes on to cover most, but not quite all, of the Empire in a somewhat random fashion (FIG. I). While six routes are given in Sardinia and one even in Corsica, there is none in Crete or Cyprus—in contrast to the Peutinger Table,3 which includes routes in Cyprus and Crete but not in Corsica and Sardinia. On the mainland the most obvious omission is that of transdanubian Dacia, but Achaea and Macedonia are also poorly represented and there are no routes at all in Lycia and Pamphylia. While Spain is treated generously, there are large gaps in Gaul, especially in Armorica; and even Spain is not dealt with as a province, or group of provinces, on its own, since some of its routes spill over into Gaul. A consideration of this and similar cases shows that the apparent coherence of the British section is largely fortuitous and derives from the fact that Britain is an island.
1 The text used in this paper is that of Cuntz, O., Itineraria Romana, i (Leipzig, 1929)Google Scholar. This has superseded the edition of Parthey, G. and Pinder, M. (Itinerarium Antonini Augusti et Hierosolymitanum, Berlin, 1848)Google Scholar, but Cuntz, like Parthey and Pinder, retains the pagination of Wesseling, P.: Vetera Romanorum Itineraria (Amsterdam, 1735), so that for reference purposes the Antonine Itinerary runs from 1.1 to 486.17 (land itinerary) and 487.1 to 529.6 (maritime itinerary).Google Scholar
1 The Roman mile consisted of 1,000 passus of five pedes each. The approximate value of the pes has been established by the measurement of it on the tombstone of T. Statilius Aper, a mensor aedificiorum in Rome (C.I.L., vi, 1975) and by the discovery of several foot-rules (two are exhibited in the Department of Greek and Roman Antiquities in the British Museum). The most commonly accepted figure is 11.65 ins. or 296 mm., giving the length of the Roman mile as 1,618 yards or 1,480 metres. The stadium, used in measuring distances at sea, was equal to one-eighth of a Roman mile, and the leuga, which was used in Gallia Comata and appears in that area in the Itinerary, was equal to 1½ Roman miles (Isidore, , Etymologiae xv, 16, 1–3).Google Scholar
3 For a convenient edition of the Peutinger Table see Miller, K., Die Peutingersche Tafel, 1916 (reprinted Stuttgart, 1962)Google Scholar; the coloured map, however, is not an exact facsimile. For a recent commentary see A., and Levi, M., Itineraria Picta (Rome, 1967).Google Scholar
4 Kubitschek, W., article ‘Itinerarien’ in Pauly-Wissowa-Kroll, , Realencyclopädie der Klassischen Altertums- wissenschaft ix (1916), 2308–63; ‘perhaps unwisely’ because the map displays some ignorance of Romano-British geography, but the article remains a fundamental work.Google Scholar
5 The relationship of the maritime to the land section is uncertain. The heading ‘Item Imperatoris Antonini Augusti Itinerarium Maritimum’ appears in all the best MSS., which also consistently link the two (see Appendix I). As is remarked below (p. 37), the maritime section does include some late first or early second-century material, but this is only an argument against a connection if one has already made up one's mind on the date (or dates) of the land itinerary. On balance there seem to be no good grounds for separating them.
6 Vegetius, Epitoma Rei Militaris iii, 6: Primum itineraria omnium regionum, in quibus bellum geritur, plenissime debet habere perscripta, ita ut locorum intervalla non solum passuum numero sed etiam viarum qualitate perdiscat, conpendia deverticula montes flumina ad fidem descripta consideret, usque eo, ut sollertiores duces itineraria provinciarum, in quibus necessitas gerebatur, non tantum adnotata sed etiam picta habuisse firmentur, ut non solum consilio mentis verum aspectu oculorum viam profecturus eligeret. (First of all, he must have itineraries of all those regions in which the war is being fought very fully set out, so that he may become fully acquainted with the distances between places in terms not only of mileage but also of the quality of the roads, and may have for his consideration reliable descriptions of the short cuts, alternative routes, mountains and rivers—in fact the more expert generals are stated to have had not only annotated but pictorial itineraries of the provinces in which they were operating, so that one could decide which way to take not only by mental calculation but by visual inspection too.) Similarly the Peutinger Table is not (as it is sometimes claimed to be) a true itinerarium pictum in the sense that Vegetius intends.
7 On the Cursus Publicus see Pflaum, H. G., ‘Essai sur le Cursus Publicus’, Mems. Acad. Inscrs. et Belles Lettres, xiv (1940), 189 ff. and 217 ff., and A. H. M. Jones, The Later Roman Empire (1964), ii, 830 ff. For an example of the equation of the Antonine Itinerary with its routes, I. A. Richmond, Roman Britain (2nd edn., 1963), 91.Google Scholar
8 For example, J. Oliver Thomson, History of Ancient Geography (1948), 376.
9 Berchem, D. van, ‘L'Annone militaire dans l'Empire Romain au IIIe siècle’, Mems. Soc. Nat. des Antiquaires de France, xxiv (=8th ser. x) (1937), 117–202 (on the itinerary specifically, 166–81).Google Scholar
10 Pliny, Epp., vi, 31. Centumcellae does, by contrast, appear as a portus in the land itinerary (300.6–7). For comment on this early route see R. Lugand, ‘Note sur l'ltinéraire Maritime de Rome à Arles’, Mélanges de I'Ecole Francaise à Rome, 1926, 124–39.
11 Byzantium is once (138.5) glossed ‘qui et Constantinopoli’, but the gloss is missing from one of the best MSS (P) and the city gets no special treatment. Similarly Cirta in Numidia (later Constantina) and Antaradus in Phoenicia (later Constantia) appear under their old names.
12 Notably Kubitschek, loc. cit., and H. Stuart Jones, Companion to Roman History (1912), 49.
13 Tierney, J. J., ‘The Map of Agrippa’, Procs. R. Irish Academy, lxiii, Sec. C., No. 4 (1963), 151–66.Google Scholar
14 Goodchild, R. G., ‘The Coast Road of Phoenicia and its Roman Milestones’, Berytus, ix (1949), 91–128.Google Scholar
15 O.S. Map of Roman Britain (3rd edn., 1956), index p. 43; R.I.B., i (1965), Nos. 2219–2314. For later additions to the former since 1955 and to the latter since 1954, see J.R.S. passim.Google Scholar
16 Tongres, C.I.L., xiii, 9158; Autun, C.I.L., xiii, 2681.
17 As, for example, on the Thurmaston milestone (R.I.B. 2244), our best civilian example, the distance is measured ‘a Ratis’ (from Leicester). For a recent comment on the system in Britain, see Graham, A. J., J.R.S., Ivi (1966), 97–98.Google Scholar
18 R.I.B., 2241, with the inscription ‘A.L.S. M.P.XIIII’ (? = A Lindo Segelocum mp xiiii), found in 1879 opposite the Lion and Snake Inn, Bailgate.
19 The normal MS. forms are iiii and viiii (not iv and ix); these are consistently used in the Itinerary, though there is some variation in the Peutinger Table.
20 As recognized by Wheeler, G. H., English Historical Review, xxxv (1920), 377–82 and xlvii (1932), 622–26.CrossRefGoogle Scholar
21 The serial numbers of the itinera do not appear in the text but have been in general use by English antiquaries since the sixteenth-century version of the British section by Robert Talbot, published by Thomas Hearne in his edition of Leland's Itinerary (2nd edn, 1744, Vol. iii, 122–64). They are obviously convenient for reference, but their use has sometimes tended to obscure the fact, noted above (p. 34), that it is only because Britain is an island that this section is so neatly grouped together; in no other province could a similar system of numbering be devised.
22 Arch., xciii (1949), 30, s.v. Corielopocarium: ‘The form CORSTOPITUM is impossible in Celtic and must be corrupt.’Google Scholar
23 Not. Dig. Occ, xl, 31. The garrison is given as numerus Supervenientium Petueriensium, and although no epigraphic confirmation is yet forthcoming the identification of Derventio as the fort at Malton (occupied from Flavian times to the fourth century) is highly probable.
24 R.I.B., 707; Birley, E. in Corder, P. and Romans, T.: Excavations at the Roman Town of Brough-Petuaria 1937 (1938), 61Google Scholar (= East Riding Arch. Soc. Trans., xxviii (1939), 229Google Scholar); cf. Richmond, I. A. and Crawford, O. G. S., Arch., xciii (1949), 31, s.v. Decuaria.Google Scholar
25 Richmond and Crawford, op. cit., 31–32, s.w.
26 ‘77–3 (in Cappadocia); 212.2 (in Cilicia); 259.13 and 260.6 (in Pannonia); 272.5 (in Dalmatia); and 398.3 (in Tarraconensis); cf. Praetorium Agrippinae (in Germania Inferior) and others in the Peutinger Table. In none of these cases does it refer to the administrative headquarters of a province, but the suggestion in J. S. Wacher: Excavations at Brough-on-Humber (1969), 26, n. 2, that praetorio should be retained here destroys the reason for attaching it to Brough. For the governor's praetorium at York see R.I.B., 662–63 and I. A. Richmond in R.C.H.M. Roman York (1962), xxxvii.
27 A small corroborative detail might be the ablative case of Derventione, as contrasted with Isurium, Eburacum and perhaps Delgovicia, but the use of cases in the Itinerary seems to be random.
28 Kubitschek, loc.cit.
29 Although the equation of Bovium with Holt cannot be taken as certain, since Holt is not on the direct road from Chester to Whitchurch; see Grimes, W. F., Y Cymmrodor, xli (1930), 6–8, with a summary of earlier theories.Google Scholar
30 Richmond and Crawford, op.cit., 43, s.v. Pampocalia.
31 Ptolemy, Geography, ii. 17.2.
32 Almondbury, the most favoured candidate, is rather far away (5 miles as the crow flies) and a smaller work, perhaps that on Old Lindley Moor (SE 091182), is more probable. It may be noted that so far Slack has yielded no material later than thesecond century (Yorks Arch.J., xlii (1966), 86–87Google Scholar. See J.R.S., lix (1969), 207.Google Scholar
33 Bede, Hist. Eccles., ii, 14.
34 Villy, F., Bradford Antiquary, viii (1940), 17Google Scholar; road not found here, Yorks. Arch. J., xli (1964), 165.Google Scholar
35 Wheeler, G. H., English Historical Review, xxxv (1920), 377–82;CrossRefGoogle ScholarCrawford, O. G. S., J.R.S., xiv (1924), 137–41.Google Scholar
36 The evidence for a Roman town at Crayford is circumstantial (V.C.H. Kent, iii (1932), 88) but considerable; it is supported by the fact that the Britons made a stand there against Hengest and Aesc (Anglo-Saxon Chronicle, anno 457).Google Scholar
37 The use of the tribal, as opposed to the proper, names for cities is, of course, quite normal in Gaul; see my discussion in J. S. Wacher (ed): The Civitas Capitals of Roman Britain (1966), 108–9. The failure to realize this, however, and the determination to find another candidate for Icinos led to an astonishing proliferation of theories, summarized by Yorke, A. C., Cambridge Antiquarian Soc. Procs., xi (1905), 2.Google Scholar
38 Rostrata Villa (124.8), 24 miles up the Via Flaminia from Rome and the first staging-point on van Berchem's main route; and seven in Africa (Villa Serviliana, 42.6; Villa Fulgurita, 59.3; Villa Magna, Villa Privata, 60.3; Casas Villa Aniciorum, 61.2; Vox Villa Repentina, 62.1; Megradi Villa Aniciorum, 62.3; Minna Villa Marsi, 63.1). Apart from Villa Magna (which is not the Villa Magna of C.I.L., viii, 25902), they were presumably, by the time of compilation, imperial estates, run by bailiffs, and this may explain why the buildings at Scole (Norfolk Archaeology, xxx (1949–1952), 151) are not luxurious. Villas are also mentioned in the Bordeaux Itinerary, but as landmarks rather than as staging-points.Google Scholar
39 Salway, P., Cambridge Antiquarian Soc. Procs., lx (1967), 39–80.Google Scholar
40 This phenomenon is clearly displayed on both the O.S. Map of Southern Britain in the Iron Age (1962) and the 0.S. Map of Roman Britain (3rd edn., 1956), where the trackway shown is the medieval Icknield Way. The later distortion was probably caused by the development of Saxon Thetford.
41 Phillips, C.W., Arch.Journ., xci (1934), 179.Google Scholar
42 This suggestion was made to me by the late Professor R. G. Goodchild.
43 Stukeley, W., An Account of Richard of Cirencester, London, 1757, 62 (Iter XVIII), Isannaria; 67, Isanna-varia, equated with Towcester.Google Scholar
44 See note 2, page 34. If the tradition of the kingdom persisted, this would be the most likely place in Britain to find this Celtic usage.
45 As Nero suppressed the Cottian kingdom in the Alps on the death of M. Iulius Cottius (Suetonius, Nero, 18).
46 Ptolemy, Geog. ii, 3, 28 (for Noviomagus see also i, 15, 6).
47 Ravenna Cosmography, v, 31 (Pinder and Parthey edn. 426, 13–14; Richmond, and Crawford, , Arch., xciii (1949), 17, No.44).Google Scholar
48 The suggestion that Regentium stands for Regnensium seems to appear first (without supporting argument) in Haverfield's article on Regni in Pauly-Wissowa, 1914, 509: ‘Navimago Regentium (wohl Noviomago Regnensium)’, whence it is repeated by Macdonald in Pauly-Wissowa, 1936, 1213, s.v. Noviomagus, and so by Richmond, and Crawford, in Arch., xciii (1949), 42Google Scholarx, s.v. Navimago Regentium. Both of the Pauly-Wissowa articles and Richmond and Crawford all quote the Antonine Itinerary as giving Regnum–although Haver-field in V.C.H. Surrey, iv (1912), 348 (to which Macdonald and Richmond and Crawford evidently intend to refer when they cite V.C.H. Surrey, i, 348) correctly quotes it as giving Regno.Google Scholar
49 For the use of the tribal name in place of the proper name, cf. Icinos in Iter V and see note 37, p. 47.
50 Found 1880 in Beancroft Road. The stone was first inscribed for Decius (R.I.B. 2273), then up-ended and re-inscribed for Gallus and Volusian (R.I.B. 2274); the mileage figure is included in the latter inscription.
51 For the argument that ‘Ad Taum’ represents ‘Ad (Ven)Ja (Icenor)um’ see Wheeler, G. H., English Historical Review, xxxv (1920), 377–82;CrossRefGoogle Scholar the point was not taken up by Crawford in his otherwise useful discussion of this Iter in J.R.S., xiv (1924), 137–41.Google Scholar
52 It would seem to be indicated by the positioning of Sinomagi in the Peutinger Table (PI. VII).
53 Ptolemy, Geog, ii, 3,6.
54 Haverfield, F., Arch. Journ., Ixxii (1915), 77–84Google Scholar, developed and amended by Collingwood, R. G., Arch., lxxi (1920–1921), 1–16.Google Scholar
55 In the Cosmography (Pinder and Parthey edn., 430, 12–15; Richmond and Crawford, 113-16) the sequence is Galluvio (=Galava) —Medibogdo (?Hardknott)—Cantiventi—Iuliocenon (?=TunnoceIum = Moresby); in the Notitia (Occ. xl), Tunnocelo—Glannibanta—Alione—Bremeten-raco.
56 Appendix II, s.v.
57 This is probably a river name (Appendix II, s.v.), but this raises no difficulty because the streams here bear the later Scandinavian names of Brathay and Rothay.
58 Appendix II, s.v.
59 See discussion in Appendix II under the place name Cunetio.
60 As Whitaker (History of Richmondshire, ii (1823), 267Google Scholar), followed by others; for a summary see Birley, E., Trans. Cumbd. and Westmd. Ant. and Arch. Soc. (NS), xlvii (1948), 4–9. The suggestion is dismissed by Haverfield (op. cit., 82) without explanation.Google Scholar
61 Collingwood, loc. cit. Haverfield had preferred Lancaster.
62 For finds at St Asaph (coins only) see Ellis Davies, Prehistoric and Roman Remains of Flintshire, 1949, 332; Flintshire Hist. Soc. Pubns., xv (1954–1955), 150, The identification has been questioned (e.g. V. E. Nash-Williams, Roman Frontier in Wales, 1954, 23), but the alternatives suggested do not carry conviction.Google Scholar
63 E.g. Nash-Williams, op. cit., 89–91; Margary, I. D., Roman Roads in Britain, ii, 1957, 59.Google Scholar
64 Margary road 60 (d).
65 This still holds good, despite the discovery of a fort partly underlying the castle at Loughor in July 1969. Hendy seems more likely than Pontardulais because the fort would then, like those at Neath and Carmarthen, be on the west side of the river.
66 Bull. Board of Celtic Studies, xiii (1949–1950), 239 ff.Google Scholar
67 Mr. A. H. A. Hogg informs me, however, that careful inspection of air photographs and ground reconnaissance have so far failed to locate it.
68 This gives the same result as that arrived at by A. D. Passmore, The Roman Road from Caerleon to Silchester, 1948. Passmore's argument, however, cannot stand, because it is based on the supposition that the distance from Gloucester to Cirencester was 14 (instead of 18) Roman miles; other figures in his table suggest that he was using a false value for the Roman mile. For recent excavations at Wanborough by Mr. E. Greenfield, see Wilts. Arch. Mag., lxiii (1968), 109–10. On the name Durocornovio, see Appx. II, s.v.; the connexion of the Cornovii with this part of the country is obscure.Google Scholar
69 See Appx. II, s.v.
70 Cf. Margary, I. D., Roman Roads in Britain, ii, 1957, 256.Google Scholar
71 So 493.12–13: Item a portu Augusti urbis traiectus in Africa Cartaginem, stadia ūccl.
72 So 138.5—139.2: Bizantio… Calcedonia, traiectus in Bithinia, m.p. iiii.
73 So 333.9—10: A Callipoli traiectum in Asia Lamsacum usque, stadia lx. This is analogous to the crossing from Gaul to Britain (not actually called traiectus) which introduces the British section (463.3—5: Iter Britanniarum. A Gessoriaco de Gallis Ritupis inportu Britanniarum, stadia numero cccl).
74 So 369.2, Utrecht, where the modern name indicates that Traiectus was, or became, the proper name; the route does not actually cross the river. On the other hand Travectus (461.9), at the crossing of the Dordogne on the road from Agen to Périgueux, seems to be identical with the Diolindum of the Peutinger Table (P. Barriére, Vesunna Petrucoriorum, 1930, 193–94, 199, identifying it as Pontours, near Lalinde).
75 Grenier, A., Archtologie Gallo-Romaine, II(i), 1934, 186, citing Cumont on Maestricht = Mosae Traiectus.Google Scholar
76 E.g. Cuntz, in his apparatus criticus.
77 V.C.H. Somerset, i, 1906, 347–48. Haverfield is in error, however, in saying that the figure between Bitton and Bath needs to be changed from vi to xi.Google Scholar
78 E. g. Haverfield, , V.C.H. Hants, i (1900), 321, with discussion.Google Scholar
79 Clarke, A., Procs. Hants Field Club, xxi (1959), 83–97.Google Scholar
80 The road running east out of Winchester which is marked as ‘course probable’ on the O.S. Map of Roman Britain, 3rd edn, 1956, is in fact a turnpike road (information from Professor C. F. C. Hawkes); but an eastward road still seems probable, perhaps on a different alignment.
81 A. Clarke, op. cit., 88.
82 See O.S. Map of Roman Britain, 3rd edn, 1956 and P. Corder (ed.), Romano-British Villas (C.B.A. Research Report, No. 1), 1955, 31–32.
83 Cf. Haverfield, , V.C.H. Hants, i (1900), 323.Google Scholar
84 Arch., xciii (1949), 25Google Scholar, s.v. Bindogladia; Wallace, W. G., Procs. Dorset Nat. Hist, and Arch. Soc, liv (1932), 87Google Scholar. Mr. H. Shortt (ibid., lxxxix (1967), 160–63) reverts to Pitt-Rivers's identification with Wood-yates, but this involves the idea of a late ‘revision’ of the Itinerary, which is not acceptable on general grounds.
85 Cf. Margary, I. D., Roman Roads in Britain, ii, 1957, 256.Google Scholar
86 E.g. Arch., xciii (1949), 41Google Scholar, s.v.; J.R.S., li (1961), 188; and, most recently, H. Shortt, op. cit.Google Scholar
87 Pinder and Parthey V, 31 (p. 425, 9–14, with reading Londinis), Richmond and Crawford 23–26. In fact the road north-east out of Exeter appears to be traversed three times, with three branches beyond, for the similarity of the three sequences can hardly be accidental, viz.:
This suggests that Alauna may be the Axe or its tributary the Yarty, which would agree well with Ptolemy's ‘αλαŨμον ποταµοŨ εκβολαι (Geog., ii, 3, 4), but would raise difficulties over the name (Axe should be Isca and Yarty does not reach the sea). Dr. S. Applebaum's identification of Alauna as the Alcester in Shaftes-bury (J.B.A.A.3, xvii (1954), 77Google Scholar) must fail, because the name here is of monastic origin (see V.C.H. Warwicks., ii (1908), 59Google Scholar, on land granted to the Abbey of Alcester in 1307; cf. Mayo, C. H., Procs. Dorset Nat. Hist, and Arch Soc., xv (1894), 43Google Scholar). Mr. H. Shortt's identification of Alauna as the Stour (op. cit.) involves the rejection of Ilchester as Lindinis.
88 Stevens, C. E., Procs. Somerset Arch, and Nat. Hist. Soc, xcvi (1952), 188.Google Scholar
89 This identification was first suggested by P. O. Hutchinson, Gents. Mag. 1849 (i), 137–46, but he later preferred Hembury (Trans. Devon. Assocn., xiv (1882), 516–24).Google Scholar
90 Lady Fox, while not agreeing with the identification, has kindly confirmed that the sea is so visible.
91 R.I.B. 2240 and 2241 (Lincoln); 2244 (Thurmaston); 2243 (Buxton); 2250 (Kenchester); 2265 (Rhiwiau-uchaf); 2272 (Caton); 2274 (Castleford); 2313 (Ingliston); and J.R.S., Iv (1965), 224 (Brougham).Google Scholar
92 R.I.B. 2235 (Castor); 2241 (Lincoln); 2244 (Thurmaston;) 2243 (Buxton); 2264 (Ty Coch); 2265 (Rhiwiau-uchaf); 2272 (Caton); 2274 (Castleford); 2283 (Middleton); 2299 (Crindledykes); 2306 (Cawfields). The late Professor Richard Goodchild suggested to me that the figure may sometimes have been added in paint.
93 R.I.B. 2247 and 2248.
94 For recent excavations at Whitchurch, see Jones, G. D. B. and Webster, P. V., Arch. Journ., cxxv (1968), 193–254Google Scholar. The unsettled nature of the area is demonstrated by the absence of villas (cf. Corder, P. (ed.): Romano-British Villas (C.B.A. Research Report, No. i), 1955, 32Google Scholar) and perhaps by the persistence of military occupation at Wall (Gould, J., Trans. Lichfield and S. Staffs Arch, and Hist. Soc. viii (1966–1967), 10–12). Mr. C. E. Stevens has suggested to me that many of the Itinerary routes in Gaul represent moves against the Bacaudae.Google Scholar
95 The overlapping of itinera may be summarized as follows:
It is difficult to draw any firm conclusions from these tables. Iter VIII appears to be travelling faster than Iter II or Iter VI; but while Iter II omits the stage Tripontio, it includes Sulloniacis, which is omitted by Iter VI. It may, however, be observed that variation in stages is a normal feature of the Itinerary, and the omission from Iter VIII of Lactodoro (Towcester), while the less important place of Magiovinio (Dropshort) is included, is especially noteworthy. The tables also give a good idea of the variation in spelling.
96 Since most towns were continuously occupied, military establishments offer the best hope. If Camulodunum in Iter II is to be identified with Slack, the absence there of material later than the 2nd century may be significant (see note 32, p. 44), but without the total excavation of the fort and any possible mansio the evidence must be inconclusive; a firm terminus ante quern for this Iter will be available when we are quite certain of the date at which Birrens was evacuated. The mention of Legio VI in Iter I places it after A.D. 122 and the absence of a Hadrianic phase at High Rochester makes it Antonine or later. But, as we have seen, these dates cannot be extended to cover other itinera.
97 It is probably idle to speculate by whom, but it would be a pleasing irony if some of Geta's journeys as governor were incorporated in a work to which his murderer's name was attached.
98 See Kubitschek, W., Wiener Studien, xiii (1891), 177–209Google Scholar; Cuntz, O., Wiener Studien, xv (1893), 260–98; and Cuntz's introduction to his edition.Google Scholar