² I am most grateful to Professor R. Stillwell and Dr. W. H. Plommer for their helpful comments on successive drafts of this paper; they are not, however, responsible for the shortcomings that remain.

³ Argive Heraeum i. 127–30, pls. 20–2 (contributed by E. L. Tilton).

⁴ A new study of the Argive Heraion as a whole is being undertaken by R. Mason of the American School of Classical Studies. This should avoid the danger, inherent in studies of the individual buildings, of dodging problems by attributing awkward blocks to another building. Since many of the blocks belonging to the South Stoa are now lying upside down and half buried, a full examination of them is impossible without lifting equipment—not available to the present author.

⁵ The need for re-examination of these features was recognized by Amandry in his important paper on the architecture of the Heraion (n. 1 above).

⁶ Since some of these have been placed in a line near the krepis of the stoa, somebody before the present author must have realized their significance.

⁷ Another fragment, similar in material and technique and 0·80 m. long (broken), 0·61 m. deep, and 0·29 m. high, may also be part of a step block of the stoa; it is now among the stones below the South Stoa.

⁸ I was able to locate ten of them, either on the krepis foundation or among the stones below the Stoa.

⁹ Recognized as such by G. Roux (Roux 11).

¹⁰ The column trace is centred on the middle of the block; the setting lines consist of a small cross near the front of the block, on the axis of the column, and a short line, also on the axis of the column, near the rear edge of the block. Similar setting lines are to be seen on several of the inner column base slabs. I am grateful to Mr. C. K. Williams for independent confirmation of the reality of the column trace.

¹¹ Argive Heraeum i. 129.

¹² Argive Heraeum i. 130; Amandry 249 n. 62.

¹³ Argive Heraeum i. pl. 22 top left; Amandry 249–50 fig. 12.

¹⁴ Amandry fig. 13.

¹⁵ Cf. the narrow triglyph where the pronaos frieze of the Temple of Zeus at Olympia turns to the long sides of the cella building (Curtius, E., Adler, F., Olympia ii (1892) 10 pl. 15Google Scholar).

¹⁶ Both frieze height and architrave height appear to have varied by up to a centimetre from block to block.

¹⁷ This is described by Tilton, (Argive Heraeum i. 130Google Scholar) as a terracotta fillet which completes the entablature, but it is a fairly normal type of eaves tile with flat decorated soffit and decorated front edge (cf. BSA lix (1964) pl. 24d). The wedge-shaped projection fitting a cutting in the cornice is unusual, but there is a similar arrangement at Assos (Bacon, J.et al., Investigations at Assos (1902–1921) 155, 167Google Scholar). The sima attributed to the stoa by Tilton, (Argive Heraeum i. 130 pl. 23GGoogle Scholar) does not belong (Shoe, L. T., The Profiles of Greek Mouldings (1936) 33 pl. 18.4Google Scholar; Amandry 248 n. 59), but the antefix (Argive Heraeum i pl. 23B) seems quite reasonable.

¹⁸ Argive Heraeum i pl. 22.

¹⁹ 0·23087 (tan 13°) x 6·803= 1·5706 m.

²⁰ Argive Heraeum i pl. 22.

²¹ Amandry 247–9. See also AJA 77 (1973) 16.

²² I was able to locate four of them, three among the blocks below the Stoa and one further west below the West Building.

²³ Besides the large cuttings for rafters described here and the wedge-shaped cuttings for the eaves tiles (see above and n. 17), the cornice blocks have H-clamp cuttings on the top face, and on each joint face two square holes c. 0·05 × 0·05 × 0·10 m. deep. Similar holes are to be found in the cornices of several Archaic and Classical buildings (Orlandos, A., Τὰ Ὑλικὰ Δομῆς τῶν Ἀρχαίων Ἑλλένων i (1958) 171–2Google Scholar), and have been interpreted either as for a lifting device or as for horizontal dowels to connect adjacent cornice blocks. In the present instance at least, the second explanation must surely be wrong for:

(a) The ‘dowel’ cutting is sometimes intercepted by a rafter cutting (Amandry pl. 67c); if the dowel was felt to be of permanent value, it could easily have been put lower down in the block.

(b) Since the cuttings are o·10 m. deep, each block would have to be set down c. 0·10 m. from its neighbour, then pushed sideways on to a dowel which, if it was to be of any use, would have to be a tight fit. The awkwardness of such a procedure explains why the Greeks used clamps, not dowels, for horizontal connections.

(c) If the blocks were already firmly connected by dowels, there would be no point in using H-clamps as well (cf. Amandry fig. 18 pl. 67b).

The lifting device was perhaps not two pairs of tongs as suggested by Orlandos (op. cit. fig. 119.10), but four simple hooks or bolts fixed to four short ropes radiating from a central lifting rope, a system which is still in use for loading and unloading ships. The square shape of the holes is presumably an indication not so much of what went into them as of how they were made—with a chisel rather than a drill.

²⁴ Bouras, Ch., Ἡ Ἀναστήλωσις τῆς Στοᾶς τῆς Βραυρῶνος (1967) 105–22.Google Scholar

²⁵ Bohn, R., Die Propyläen der Akropolis zu Athen (1882) pl. 14.9.Google Scholar

²⁶ Hodge, A. T., The Woodwork of Greek Roofs (1960) 48.Google Scholar

²⁷ Hesperia vi (1937) 36.

²⁸ e.g. Nike, temple (Hesperia xxxii (1963) 378Google Scholar), Ionic columns from the Agora, (Hesperia xxix (1960) 351–4).Google Scholar

²⁹ Hesperia xl (1971) 276–8.

³⁰ The inner column height restored by Thompson, 6·174 m., would therefore seem to be incompatible with the more complex type of roof.

³¹ Hesperia xl (1971) 243–55.

³² Fabricius, Doerpfeld, and Choisy envisage a roof of rafters carried on inner architraves and a ridge-beam, while Marstrand and Jeppersen (whose discussion in Paradeigmata (1958) 69–101 is the best recent one) restore closely spaced purlins carried by principal rafters, with no ridge-beam or common rafters. The most natural meaning of the terms used in the inscription would seem to be that required by the restoration of Fabricius-Doerpfeld-Choisy, although looseness of usage and lack of certain definitions make the linguistic argument an inconclusive one. The fact that the and the would be equally stressed in his version of the roof, but not in the other, is taken by Marstrand as an argument in its favour. In his version, however, the inner architraves would carry no load at all, while in the other version at least all the beams are doing some work, so that this line of argument would seem to favour the Fabricius-Doerpfeld-Choisy roof as well. Finally, normal Greek roofing practice argues against the omission of a ridge beam as in the Marstrand-Jeppersen restoration (Hodge, A. T., The Woodwork of Greek Roofs (1960) 45Google Scholar), and although roofs without common rafters are known (Hodge, op. cit. 18–20, 26–32, 52, 72–5) they are less common than those with rafters.

³³ Courby, F., Délos v (1912) 35–6.Google Scholar

³⁴ BCH lxxxvi (1962) 3°2.

³⁵ Humann, C., Magnesia am Mäander (1904) 133–4.Google Scholar

³⁶ Jdl xxxi (1916) 306–9.

³⁷ That is supposing that the spacing of the sloping cross-beams was half the inner intercolumniation and that they carried two purlins. We must assume of course that the weight of tiling and battens, etc., was the same in both cases and that the timber was stressed to the same degree in all beams.

³⁸ BSA lxiii (1968) 165–9.

³⁹ A. T. Hodge, op. cit. (n. 32) 45.

⁴⁰ This may seem extraordinarily large for a beam spanning only 4·57 m. but the Arsenal of Philo was to have inner architraves c 0·80 × 0·72 ( ft. high) spanning only c. 3·60 m. ( ft.) (Jeppersen, K., Paradeigmata (1958) 72, 90).Google Scholar

⁴¹ Olympia, Temple of Zeus 0·765

Athens, Temple of Hephaistos 0·756

Athens, Parthenon 0·776

Sounion, Temple of Poseidon 0·774

Rhamnous, Temple of Nemesis 0·761

Athens, Stoa of Zeus 0·762

⁴² Temple of Zeus at Olympia, c. 0·049 m. per metre; buildings of Periklean Athens 0·0375–0·44 m. per metre. Since the relation of upper to lower diameter remained roughly constant, the rate of taper naturally decreased when the column height was increased in Periklean Athens.

⁴³ That is, assuming there was no entasis. The inner columns apparently do have entasis, however, the rate of taper varying from 0·028 m. per metre at the bottom to c 0·042 m. per metre at the top (Argive Heraeum i. 128) so that the outer columns probably did, too; but that would not materially affect the point at issue here.

⁴⁴ The entablature of the Nike Temple is nearly the same height as that of the temple by the Ilissos, although its columns are 0·429 m. lower. So too the entablatures of the Hephaisteion, the Temple of Poseidon at Sounion, and the Temple of Ares at Athens are almost identical in height although their columns are 5·713, 6·024, and c. 6·225 m. high respectively. The columns of the east and west porticoes of the Propylaia differ by about 0·30 m. in height but carry the same entablature and have the same lower diameter. The inner columns of the Stoa at Oropos all have the same lower diameter but four were probably c. 0·50 m. lower than the others (BSA lxiii (1968) 167), and a similar set of columns comes from the Agora, Athenian (Hesperia xxix (1960) 351–4).Google Scholar In all the peripteral temples of mainland Greece in the Archaic period the columns on the flanks had the same height as those on the front, but a different lower diameter. Not all these instances occur for the same reason, but at least they show how illusory is the idea that Greek architects adhered rigidly to an accepted system of proportion. In fact if we assume that they were to a greater or lesser extent consciously experimenting with the proportions of the Doric order, rather than mindlessly applying a traditional system (which mysteriously evolved and developed in the national subconscious?), then one of the most natural and logical ways for an architect to proceed to try out a new idea would be to keep every other dimension the same as in his last building, but to increase (for instance) the column height or reduce the lower diameter.

⁴⁵ Koldewey, R., Puchstein, O., Die Griechischen Tempel in Unteritalien und Sicilien i (1899) 27.Google Scholar Cf. Bundgaard, J., Mnesicles (1957) 144–5Google Scholar fig. 49.

⁴⁶ An arrangement of this sort is suggested for the Stoa Basileios at Athens (Hesperia xl (1971) 244), in view of the small lower diameter of the inner columns.

⁴⁷ Hesperia xxxix (1970) 244–7.

⁴⁸ e.g. Stoa at Oropos, South Stoa at Corinth, North Stoa at Priene, etc.

⁴⁹ De arch. v. 9. 3.

⁵⁰ Figures taken from the Chronological List of Temples at the back of Dinsmoor.

⁵¹ See the tables of proportions given by Amandry (Amandry 257) and enlarged by Roux (Roux 410–11), and byCh. Bouras (op. cit. (n. 24) 150–3).

⁵² Roux, passim, emphasizes the Peloponnesian character of the temple at Bassai. But it stands at the head of the Peloponnesian tradition that he so illuminatingly distinguishes; the tradition cannot be traced back into the first half of the fifth century or earlier in the Peloponnese. The Ionic capitals have their closest antecedents in Attica (Roux 342–5), and both the over-all proportions of the Doric order (see Dinsmoor's List of Temples, above, n. 50) and the treatment of the Doric capital (Roux 410–11, esp. Tables 3–6) are closer to Athenian practice than to the Temple of Zeus at Olympia.

⁵³ Paus. ii. 17. 3. The Athenian characteristics of the design are well brought out by G. Roux (Roux 57–62).

⁵⁴ As shown by Amandry (Amandry 269) in spite of Tilton, (Argive Heraeum i. 128Google Scholar).

⁵⁵ e.g. those given by Amandry, Roux, and Bouras (n. 51 above).

⁵⁶ See, for example, Ch. Bouras, op. cit. (n. 24) 153, 156, Tables H and IA.

⁵⁷ An excellent beginning was made with G. Roux's L'Architecture de l'Argolide (1961).

⁵⁸ The proportion column height/entablature height keeps close to these simple proportions with remarkable consistency within the three groups, and must therefore have been an important proportion to the architects. That should not surprise us, for the relation of solid entablature to open colonnade is one of the most important factors governing the impact of a Doric façade.

⁵⁹ Mainland architects seem to have preferred a column height times the axial intercolumniation in the sixth century, then changed to a proportion of 2 times the axial intercolumniation in the fifth, beginning with the Temple of Aphaia at Aigina. Sicilian and Italian architects were less consistent in this respect, but they too seem to have been affected by the double-square proportion in the mid fifth century. Vitruvius returns to the double square in his rules for a Doric stoa (De arch. v. 9. 3), but not for the Doric temple.

⁶⁰ Cf. Plommer, W. H., Ancient and Classical Architecture (1956) 132.Google Scholar

⁶¹ Movement of Athenian-trained masons away from Attica, as the Periklean building programme slowed down in the last quarter of the fifth century, may also have played a significant part.

⁶² Amandry (Amandry 273) would date the South Stoa to c. 450–440 B.C., a date which may well be right; but it does not seem possible to the present author to exclude a date as much as twenty years later.

⁶³ See above, pp. 69–71.

⁶⁴ It may seem capricious, after what has been said above, even to suggest using a change in proportion as an aid to dating. The point made above, however, is not that nothing changes with time, but that the change will not necessarily be gradual and continuous and that place as well as date (and possibly other factors as well) may influence the proportions of a Doric design.

⁶⁵ Courby, F., Délos xii (1931) 121–2.Google Scholar

⁶⁶ Ch. Bouras, op. cit. (n. 24) 155.

⁶⁷ Thuc. iv. 133.

⁶⁸ De arch. iii. 3. 3.

⁶⁹ The Western Greek colonies seem less concerned with this problem; Temples D and FS at Selinous, although large, have slender, widely spaced columns (intercolumniation = 2·57 and 2·50 diameters, respectively), while the small columns of the Basilica and Temple of Ceres at Paestum are closely spaced (intercolumniation = 2·00 and 2·07 diameters respectively).

⁷⁰ Column 3 shows the actual diameter, while column 5 shows what it would be if a rule D = I/2·3 were consistently applied.

⁷¹ Column 6 shows the actual intercolumnar space, while column 7 shows what it would be if D = I/2·3. The procedure amounts to a less uniform application of Dinsmoor, 's observation (Hesperia ix (1940) 22)Google Scholar that in Periklean temples the axial intercolumniation exceeds twice the lower column diameter by c. 0·50 m. Column 8 of Table 2 shows that there is no standard difference between the axial intercolumniation and twice the lower diameter in the buildings cited there. Cf. also Atti del VII Cong. Int. Arch. Class. (1961) i. 362–8.

⁷² The figures are based on those given by Dinsmoor (Dinsmoor 337–9), with some additions.

⁷³ Similarly, it was the desire to have usable intercolumniations without the expense of large columns whichled to the adoption of three-metope, and eventually fourmetope, spans in later Doric.

⁷⁴ Cf. the Stoa of Antigonos at Delos (Courby, F., Délos v (1912)Google Scholar fig. 24 pl. 3).

⁷⁵ Cf. the restoration of the Stoa Basileios at Athens (Hesperia xl (1971) 247 fig. 3).Google Scholar

⁷⁶ The frieze height, which in both the South Stoa and the Temple of Zeus cannot be simply expressed in terms of the foot used (cf. below, p. 83), seems to have been designed as one-third of the intercolumniation; in the South Stoa = 0·763 m. agreeing with the actual frieze height of 0·76–0·77 m., while in the Temple of Zeus = 1·742m., agreeing with the actual frieze height of 1·74 m. The architrave height was perhaps designed as of the lower diameter, rather than as amore complex fraction of the column height; in the South Stoa 0·88 = 0·704 m., agreeing with the actual architrave height of 0·706 m., while in the Temple of Zeus 2·21 (flank diameter) = 1·768 m., agreeing with the actual architrave height of 1·75–1·77 m. (Curtius, E., Adler, F., Olympia, Text ii (1892) 7Google Scholar, although pl. 14 gives the height as 1·78 m.).

⁷⁷ Argive Heraeum i. 129.

⁷⁸ BCH lxxx ( 1956) 112 n. 4. He suggests the unit for the foundation blocks of the krepis only, and does not press it to fit the whole building, admitting that these blocks may have been taken from another building.

⁷⁹ i.e. the number of feet and sixteenths of a foot of the size specified which corresponds most closely with the actual measurements in metres.

⁸⁰ i.e. the difference between the actual measurement and the theoretical equivalent in feet and sixteenths of the size specified. The + sign means that the theoretical equivalent is greater than the actual measurement, the — sign that it is less. These signs are ignored in the Total of Differences given in the last line of the table.

⁸¹ The depth from the inner face of the rear wall to the front of die krepis foundations is 12·74 m. (Argive Heraeum i. pl. 20). As suggested above (p. 69), the thickness of the rear wall was probably about equal to the tread width of the two steps, so the external depth of the stoa at frieze height was probably about 12–70 m. This would give figures of 40 feet (+ 0·012), feet (+0·005), and 39 feet (+0·014) for the three different units, and does not significantly affect the comparisons made below.

⁸² i.e. length and breadth.

⁸³ Cf. Dinsmoor 152, etc.

⁸⁴ i.e. the number of feet and sixteenths of a foot of the size specified which corresponds most closely with the actual measurements in metres.

⁸⁵ i.e. the difference between the actual measurements and the theoretical equivalent in feet of the size specified.

⁸⁶ For the Temple of Zeus, the dimensions from the flanks are given here. This is not so much because the flank dimensions fit the foot better as because they correspond more closely to the simple proportional system.

⁸⁷ As pointed out by Riemann, H. (AA 1952, 14–15Google Scholar, quoting Moe, C. J., Numeri di Vitruvio (1945) 98)Google Scholar, Vitruvius gives the triglyph as one possible basis for a system of proportions (De arch. i. 2. 4), and in fact in his rules for the Doric order it is the triglyph width, not the lower diameter, which is equal to the module. It remains to be shown, however, that this was also the basis of fifth-century Doric design.

⁸⁸ For the operation of this process in triglyph design see Dinsmoor's discussion of the Architect, Theseion (Hesperia ix (1940) 45–6)Google Scholar, and for its more general application, Scranton, R. L. in The Muses at Work (ed. Roebuck, C., 1969) 8–9.Google Scholar

⁸⁹ This is a disturbing difference; it may have something to do with the way of setting out or working a columns drum. At any rate it is worth noting that the front column diameter of the Temple of Zeus is 0·008 m. more than feet and that the column diameter of the South Stoa is 0·006 m. more than feet.

⁹⁰ Other approaches have usually been to assume the use of ratios which are geometrically simple but not precisely expressible in numbers (such as 1: √2); for instance the Dynamic Symmetry of J. Hambridge, The Parthenon and Other Greek Temples, Their Dynamic Symmetry (1924), the R rectangles of C. Tiberi, Mnesicle Architetto (1964), and the higher mathematics of Bousquet, J., Fouilles de Delphes, Le Trésor de Cyrène (1952) 77–98.Google Scholar In the light of what evidence we do have, the onus of proof must rest with the proponents of such systems, and to show that such a theoretical system fits a building with reasonable precision does not constitute proof. Rhys Carpenter's telling criticism of Dynamic Symmetry (AJA xxv (1921) 16–36)Google Scholar is still a vital contribution to this field.

⁹¹ IG ii². 1668. The most recent full discussion is by Jeppersen, K., Paradeigmata (1958) 69–101Google Scholar, giving text, English translation, and commentary.

⁹² The over-all dimensions, the central and side aisle widths and the pillar height are all given in fives of feet, and the other dimensions are all in half and quarter feet. Eighths and sixteenths are specified only for the woodwork.

⁹³ It may be that the kind of adjustment discussed above (p. 84) is what Vitruvius means by ‘commensus ratiocinationibus explicati’ (De arch. vi. 2. 1). This is certainly something different from the modification of the design to suit the site that he goes on to talk about.

⁹⁴ Dinsmoor admits as architecturally significant only a ‘Doric’ foot of 0·3265 m. and an Ionic foot of 0·294 m. (Dinsmoor 161 n. 1, 195 n. 1, 199 n. 3, 222 n. 2, 229 n. 2; Atti del VII Cong. Int. Arch. Class. (1961) i. 360).