¹ Dinsmoor, W. B., The Architecture of Ancient Greece (3rd ed., 1950) 173–4Google Scholar; Orlandos, A. K., Τὰ Ὑλικὰ Δομῆςτῶν Ἀρχαίων Ἐλλήνων 2 (1958) 101–16, 163–75Google Scholar; Martin, R., Manuel d'Architecture Grècque 1: Matériaux et Techniques (1965) 201–19Google Scholar. These works are cited below by their authors' name only. In addition, the following abbreviations are used:

FD Ecole française d'Athènes, Fouilles de Delphes.

KP Koldewey, R., Puchstein, O., Griechische Tempel in Unteritalien und Sicilien (1899)Google Scholar.

² Vitruvius 10.2; Hero, Mechanica 3.2–5 (Teubner, ed. Nix-Schmidt; only the first of these sections survives in Greek, the rest in an Arabic translation).

³ Pliny, , Nat. Hist. 36.14Google Scholar. The story is rejected outright by Orlandos, 101–3.

⁴ JCuneifS 7 (1953) 5–7, fig. 1; Drachmann, A. G., The Mechanical Technology of Greek and Roman Antiquity (1963) 203Google Scholar.

⁵ [Aristotle] Mech. 18 (=853a32–853b13); Drachmann, A. G., The Mechanical Technology of Greek and Roman Antiquity (1963) 15Google Scholar.

⁶ Pulleys (in the plural) are mentioned in connection with lifting machines in a fourth century B.C. architectural inscription (IG ii² 1672.156 (329/8 B.C.)). In theory these could be simple pulleys used in parallel, not compound pulley systems.

⁷ It is important to distinguish the U-shaped holes discussed in the following paragraphs (FIG. 1a) from the U-shaped channels discussed later (FIO. 1b). They are characteristic of different periods of Greek architecture, and are normally related in different ways to the centre of gravity of the blocks concerned. See note 37 below.

⁸ For lists of the buildings where U-shaped holes occur see Orlandos 165–8, Martin 210 n. 2. Correct the reference in both for the cornice of the Peisistratean temple at Athens to Wiegand, T., Die Archaische Porosarchitektur der Akropolis zu Athen (1904) 121Google Scholar, where the further reference should read Penrose, F. C., Principles of Athenian Architecture (2nd ed., 1888) pl. 46Google Scholar. Add to the lists: the early temple of Apollo at Kyrene (Pernier, L., Il Tempio e ľAltare di Apollo a Cirene (1935) 54, fig. 27–8, pl. 4Google Scholar); an archaic capital from the Akropolis at Athens (Durm, J., Die Baukunst der Griechen (3rd ed., 1910) 98, fig. 71Google Scholar); the early temple of Aphaia at Aigina (Furtwaengler, A., Aegina (1906) 140, fig. 113Google Scholar); a cornice and tympanon block from Kalydon (Dyggve, E., Das Laphrion von Kalydon (1948) 110–15, 117–18Google Scholar).

⁹ U-shaped holes in these positions in ordinary wall blocks are shown by Orlandos (fig. 119.12) followed by Martin (fig. 88), and by Dinsmoor (fig. 63) respectively. The difficulty was noticed by Bourget, E. in BCH 36 (1912) 650Google Scholar.

¹⁰ In the Heraion at Olympia, the early temples of Apollo and Athena at Delphi, the Treasury of the Corinthians at Delphi (BCH 36 (1912) 650, fig. 3), the early temple of Apollo at Kyrene and the temple of Artemis at Kerkyra. For references see Orlandos 168, Martin 210, n. 2 and above, note 8.

¹¹ This function was later fulfilled by small slots cut to take an iron crow-bar (Orlandos 129–30, fig. 70, Martin 235–6, figs. 110–11).

¹² Curtius, E., Adler, F., Olympia, Architecture (1892) pl. 18Google Scholar.

¹³ Treasury of the Corinthians at Delphi, early temple of Aphaia at Aigina, early temple at Mykenai, Peisistratid temple at Athens (Penrose, F. C., Principles of Athenian Architecture (2nd ed. 1888) pl. 46Google Scholar), West Building at the Argive Heraion, Kalydon (unattributed). For references see Orlandos 168, Martin 210, n. 2, and above note 8.

¹⁴ FD, Demangel, R., Daux, G., Le Sanctuaire d'Athena Pronaia 1 (1923) 29–33Google Scholar.

¹⁵ For this method of carrying heavy stones see Naville, E., Bubastis (1891) pl. 30Google Scholar, and cf. [Aristotle], Mech. 29 (=857b9-20). Usher, A. P., A History of Mechanical Invention (2nd ed., 1954) 157Google Scholar, gives the load carried by a man as 90 lb = 41 kg, but that is for a full day. According to Smith, J., The Panorama of Science and Art 1 (1815) 344Google Scholar, a porter used then to carry 180 lb = 82 kg on his shoulder, while a coalheaver would carry up to 250 lb = 113 kg over a short distance; Hero takes as the standard power input for his baroulkos a man who can lift 5 talents = c. 130 kg (Hero, , Mech. 1.1 (ed. Nix-Schmidt, 4, lines 3–5)Google Scholar). Four men could therefore carry 450 kg or so by means of poles. The approximate weights given here and elsewhere in this paper are based on a weight of 2¼ tons/m³ for limestone and 2¾ tons/m³ for marble.

¹⁶ FD, Audiat, J., Le Trésor des Athéniens (1933) 34, 52Google Scholar.

¹⁷ FD, Audiat, J., Le Trésor des Athéniens (1933) 52Google Scholar, followed by Orlandos 168 and Martin 210 n. 3. FD, J. Audiat, op. cit., pl. R shows, however, no hole in the southern front architrave block to match the hole near the north end of its backer.

¹⁸ Furtwaengler, A., Aegina (1906) 50, pl. 36Google Scholar. Widespread use of a crane or hoist is virtually certain from the late sixth century B.C. onwards; see below pp. 7–8.

¹⁹ There seems to be no ancient authority for calling these bosses ancones.

²⁰ Orlandos 163–5, fig. 119.1–2; Martin 209–10, fig. 86; Plommer, W. H., Ancient and Classical Architecture (1956) 150, 154Google Scholar. W. B. Dinsmoor prefers lifting tongs to loops of rope (Dinsmoor 173); the first argument used here does not then apply, but the second and third do, and the fourth applies with increased force. Earlier writers were more cautious about the purpose of the bosses: Choisy, A., ĽArt de Bâtir chez les Romains (1873) 111Google Scholar; Perrot, G., Chipiez, C., Histoire de ľArt 7 (1898) 334, 519, KP 225Google Scholar, Durm, J., Die Baukunst der Griechen (3rd ed., 1910) 147Google Scholar etc.

²¹ Ist. Mitt. 13/14 (1963–4) 32, fig. 5–7; Hogarth, D. G., British Museum: Excavations at Ephesos; the Archaic Artemisia (1908) 257, fig. 67Google Scholar.

²² Good examples of bosses with unsuitable shape or inadequate projection can be seen in Orlandos fig. 113, Martin pl. 14.2, 16.2, 17.1–2, 18.1, 27.1, 31.2, 36.2, 52.2. One set of bosses which perhaps project enough to hold loops of rope is on the drums prepared for the earlier Parthenon (JDAI 55 (1940) 242–261); some of the knobs are undercut as if to prevent a loop of rope from slipping (as drum 32, ibid. 257 fig. 9), but even here there appear to be some inadequate bosses (as on drum 44 (ibid. 248, fig. 2)). Many, but not all, of the krepis blocks of the temple at Segesta have strongly projecting bosses, but on these see below and notes 23–4.

²³ KP pl. 19; Bacon, J., Clark, F., Koldewey, R., Investigations at Assos 1881–3 (1902–1921) 141Google Scholar. Bosses also occur on corner blocks of the Propylaia at Athens and the temple of Zeus at Stratos.

²⁴ Martin 193–4. This objection in fact applies to most of the krepis blocks of the temples at Assos and Segesta, since there is a row of backing blocks tight up against the facing blocks.

²⁵ For such a use of rollers, see Hero, , Mechanica 3.2Google Scholar.

²⁶ Orlandos 171, fig. 125–6, Martin 215–16, 235. Work was usually begun at the two ends of a wall so that two teams of masons could be supplied by a single lifting device set up opposite the middle of the wall. Cf. also below pp. 6–7.

²⁷ The bosses are set low on the orthostates of the temple of Nemesis at Rhamnous, (BCH 48 (1924) 312, fig. 4)Google Scholar, the Mausoleum at Belevi (Martin pl. 17.1), and in several instances on the Propylaia at Athens (AJA 8 (1904) 43, fig. 2), the temple of Apollo at Didyma (Wiegand, T., Didyma 1 (1941) pl. 89, 136Google Scholar), and elsewhere.

²⁸ Naxos: AM 49 (1924) 17–22; AA 1968, 693–717; AA 1970, 144–52. Paros: AA 1923–4, 278–94; AM 49 (1924) 22–5; AA 1970, 144–52. A similar situation occurs in the outer column shafts of the Tower of the Winds at Athens. Being monolithic, they were fluted before being set in place, and a small boss was left in four of the flutes, a few centimetres from the ground, to allow the shaft to be positioned exactly (Durm, J., Die Baukunst der Griechen (3rd ed., 1910) 157, fig. 131Google Scholar).

²⁹ For the difficulty of removing rollers from beneath a heavy block, see AA 1968, 703, n. 8. It is noteworthy that no bosses occur on walk with a quarry-faced outer face (e.g. Martin pl. 42–3). Since it is hard to believe that a series of bosses could have been dressed off such a surface without leaving a trace, there is reason to suppose that the rough quarry face took the place of the bosses. This it could do if the bosses were intended to provide purchase for crowbars, but not if the bosses were intended to take loops of rope.

³⁰ Vitruvius 10.2.10; Hero, , Mechanica 3.2Google Scholar.

³¹ Hero, , Mechanica 3.3Google Scholar.

³² Hero, , Mechanica 3.4–5Google Scholar.

³³ See above p. 5 and note 26.

³⁴ Cf. Choisy, A., ĽArt de Bâtir chez les Romains (1873) 117–18Google Scholar.

³⁵ For lists of examples with references see Orlandos 170–2, 172–5, Martin 215–16, 218–19. The West Building at the Argive Heraion should be omitted from Martin's list for lifting tongs (Martin 215), for the relevant cuttings are U-shaped holes (Hesperia 21 (1952) 245, fig. 10). It has recently been argued that die blocks in which these cuttings occur do not belong to the West Building (AJA 77 (1973) 11–16); they must nevertheless date from the sixth century.

³⁶ See note 6 above.

³⁷ See FIG. 1. The lists of examples given by Orlandos 169 and Martin 210, n. 4 confuse the two types of cutting. Except for a block attributed to the Treasury of the Sikyonians (see below and note 39), the early buildings at Delphi have U-shaped holes, not U-shaped channels, and should be omitted from the lists. The broad, shallow grooves in the concealed long face of each of the architrave backers of Temple GT at Selinous (KP 125) may have been used to adjust them against the outer blocks by means of a lever. The backers weigh c. 40 tons and the cuttings imply a single point of suspension.

³⁸ There are some exceptions (KP 225).

³⁹ FD, Demangel, R., Daux, G., Le Sanctuaire d' Athena Pronaia 1 (1923) fig. 36Google Scholar. The building is not fully published, and it is unclear how firmly the block is attributed and whether there was a similar cutting at its other end.

⁴⁰ KP 99 record three different types of cutting, of which that referred to here is the most complex; it occurs on 3 blocks of the 12 listed by KP. The other types, occurring on 4 of the 12 blocks, have no rope groove, and so would not allow a block to be set tight against its neighbour. The remaining 5 blocks have no cutting.

⁴¹ This was done with the front architrave blocks of the temple of Poseidon at Sounion; the backers there had to be lifted with lewis irons so as to be set tight up against the front blocks (Orlandos 163, 170; for a less likely method, see BSA 45 (1950) 85), but in Temple C at Selinous the main part of the architrave consists of just a single row of blocks.

⁴² KP 105, 225, illustrated by Durm, J., Die Baukunst der Griechen (3rd ed., 1910) fig. 237Google Scholar. These cornice blocks are attributed to Temple D by Orlandos 169, followed by Martin 212–13, but KP 105 refer to Durm's illustration (in the second edition (1892) of Die Baukunst der Griechen, 117, fig. 89) in their description of Temple C, and they make no mention of such grooves in the cornice of Temple D. Only two or three of the cornice blocks have these cuttings, and KP 225 suggest they may have been among the last laid blocks in the course.

⁴³ KP 107, fig. 85.

⁴⁴ Hesperia 24 (1955) 153–7 (Corinth); Broneer, O., Isthmia 1 (1971) 13Google Scholar. Roebuck, M. C. (Hesperia 24 (1955) 156)Google Scholar comments on the lightness of the blocks from Corinth, and suggests that the rope loops were used for the general handling of the blocks, not specifically for lifting.

⁴⁵ Bacon, J., Clark, F., Koldewey, R., Investigations at Assos 1881–3 (1902–1921) 155, fig. 6Google Scholar; Orlandos 169, fig. 122. Orlandos refers also to lewis holes in this temple (Orlandos 173, fig. 127), but they occur only in the ceiling beams attributed to the temple by Clark, but justly rejected by Bacon, and Koldewey, (Investigations at Assos 166–7)Google Scholar.

⁴⁶ KP 17. I owe this observation to W. H. Plommer.

⁴⁷ So Lawrence, A. W., Greek Architecture (1957) 119Google Scholar. Robertson, D. S., Greek and Roman Architecture (2nd ed., 1945) 84, 325Google Scholar had suggested c. 560 B.C. and W. B. Dinsmoor (Dinsmoor 88) c. 540 B.C. The early date is supported by Plommer, W. H. in BSA 65 (1970) 186 n. 9Google Scholar.

⁴⁸ The roof terra-cottas of the temple (MonAnt 43 (1956) 303–9) must be later than its frieze-backers, but are hard to date. They are similar in type to those of the ‘Tavole Paladine’ at Metapontum (ibid. 309–14), which is usually dated in the late sixth century.

⁴⁹ Cf. Drerup, H., Griechische Baukunst in geometrischer Zeit (Arch. Hom. Kap. O, 1969) 106Google Scholar. Where larger stones were used, they occur in positions where they could be levered into place with crowbars, without actual lifting; cf. Cambitoglou, A. et al. , Zagora 1 (1971) 22–3, n. 7Google Scholar.

⁵⁰ For the dimensions of the block see Annuario 1 (1914) 53. Larger blocks weighing c. 1.7 tons were used in the mid-seventh century fortifications at Leontinoi (Winter, F. G., Greek Fortifications (1971) 128–9Google Scholar).

⁵¹ The intercolumniation was probably 3·07 m, the architrave width c. 1·20 m, its height unknown. If we follow the reconstruction proposed by H. Schleif (Rodenwaldt, G., Kerkyra 1 (1940) 33, fig. 92Google Scholar), the main (lower) block, c. 0·70 m high, would weigh c. 6 tons. If, as seems more likely in view of the height of the existing fragment and the position of the U-shaped hole, the architrave consisted of two rows of blocks side by side, each block would weigh about 5¾ tons.

⁵² The weight of the Sounion kouros can be estimated as c. 2 tons by taking its height as 3·10 m, and its average cross-section as c. 0·40 × 0·60 m. Since the Colossos of Delos was about four times life size (Richter, G. M. A., Kouroi (2nd ed., 1960) 51Google Scholar), its height should have been c. 7 m or 2¼ times that of the Sounion kouros; its weight would therefore be (2¼)³ = c. 11·4 times that of the Sounion kouros, or about 23 tons. It need hardly be said that this is a very approximate estimate, but it may not be an over-estimate, for Cyriac of Ancona, who apparently saw the Colossos in a much more complete state seems to give its height as 24 cubits, or 10–11 m (Archaeology 25 (1972) 213).

⁵³ Loud, G., Altmann, C. B., Khorsabad 2 (1938) 15Google Scholar.

⁵⁴ JCuneifS 7 (1953) 5–7, fig. 1.

⁵⁵ Layard, A. H., Nineveh and Babylon (Abridged; 1867) 18–28Google Scholar.

⁵⁶ Cook, J. M., The Greeks in Ionia and the East (1962) pl. 49–50Google Scholar.

⁵⁷ This has often been argued for sculpture (e.g. Richter, G. M. A., Kouroi (2nd ed., 1960) 2Google Scholar) and in general, although not in detail, for Greek architecture (e.g. Boardman, J., The Greeks Overseas (1964) 159–60Google Scholar).

⁵⁸ Clarke, S., Engelbach, R., Ancient Egyptian Masonry (1930) 86Google Scholar. This is still the most useful book on Egyptian building techniques, and the remarks below are based on it.

⁵⁹ A model rocker comes from a foundation deposit at Deir el Bahari (Naville, E., Deir el Bahari 6 (1908) 9, pl. 168)Google Scholar. The use of rockers for lifting is strongly-upheld by Choisy, A., ĽArt de Bâtir chez les Egyptiens (1904) 80–93CrossRefGoogle Scholar, but is considered of little importance by Clarke, S., Engelbach, R., Ancient Egyptian Masonry (1930) 94Google Scholar.

⁶⁰ Herodotos 2.125.

⁶¹ Pliny, , Nat. Hist. 36.14, 21, 95–6Google Scholar. Pliny, and Vitruvius, (De Arch. 7, pref. 16)Google Scholar confuse the two temples of Artemis at Ephesos. The clearest indication that Chersiphron was connected with the archaic temple is the similarity of the methods of transport developed by him and his son Metagencs (Vitr. De Arch. 10.2.11–12) to those used in the archaic Temples F and GT at Selinous (KP 119–20, 125).

⁶² It is the weight not the method of raising it that Pliny remarks on, and it was not the raising of the blocks but setting them in place that caused Chersiphron most anxiety. If the blocks could be raised by a winch-driven crane, setting them would present little difficulty.

⁶³ See the table below (p. 17). I assume that the architrave, which the extant capitals show to have been about 1· 05 m. wide, had a height about equal to the lower column diameter, and was made of a single block of marble over each span, giving a weight of about 41 tons for the central front architrave.

⁶⁴ It would presumably go out of use on the introduction of the hoist in the late sixth century. Hero-dotos was born shortly before 480 B.C.

⁶⁵ The polyspaston has 6 pulleys in each system (not counting the guide pulley at the base of the jib, of course) Vitr. 10.2.8–9. This practical limitation is naturally disregarded in the theoretical parts of Hero, 's Mechanica (Mech. 2.3)Google Scholar.

⁶⁶ Edwards, I. E. S., The Pyramids of Egypt (2nd ed., 1961) 271Google Scholar.

⁶⁷ Or animals (Hero, , Mechanica 3.3)Google Scholar. Hero seems to expect that the power will be applied directly like this, but Vitruvius directs it only for one machine (Vitr. 10.2.8–10).

⁶⁸ Hence Vitruvius emphasises that only experts can operate the polyspaston, worked by direct pulling without a winch (Vitr. 10.2.8).

⁶⁹ In contrast to Hero, Vitruvius clearly regards a winch as normal. The winch of the trispaston certainly had a horizontal axle (Vitr. 10.2.2), and the terms of his description suggest that the other devices did too. Some modern authorities prefer to show a capstan turning on a vertical axle (e.g. Dunn, J., Die Baukunst der Griechen (3rd ed., 1910), fig. 68)Google Scholar, basing the restoration presumably on the illustration to Hero, Mechanica 3.2Google Scholar (Teubner ed. Nix-Schmidt, figs. 47, 76; Drachmann, A. G., The Mechanical Technology of Greek and Roman Antiquity (1963), fig. 35)Google Scholar. A capstan can have longer bars providing greater mechanical advantage and allowing more men to exert their strength usefully, but it is considerably more difficult to provide effective bearings for a capstan, and also to brace it firmly from the crane, so that winding rope on to the capstan lifts the load rather than uprooting the capstan. The earliest reference to a winch is in Herodotos 7.36, where winches are used to tighten the cables for the bridge across the Hellespont in 480 B.C., but winches may have been used in Assyria from the seventh century B.C. (JCuneifS 7 (1953) 15–17). Winch and pulley hoists are regarded as normal for architectural use in [Aristotle], Mech. 18 (=853b10-13).

⁷⁰ If each man turns one handle continuously through a full circle, there is a weak point at the top and bottom of the turn which must be arranged to coincide with the strong point of another man's turn. If each man changes his grip from one hand spike to another in the course of the turn, so as to exert his force more effectively, then the load must be held by half the men working the winch while the other half change their grip. Figures given for the force which can be exerted by a man turning a handle vary (see the references quoted in note 15), but seem to be based on the assumption that he is working steadily all day. The figure used here, 50 kg. per pair of men, is based on the assumption that if they are working over a much shorter period and do not have to turn the winch fast, a pair of men can exert at every point in the turn a force nearly equal to the full weight of one man.

⁷¹ Vitruvius 10.2.5–7.

⁷² Drachmann, A. G., The Mechanical Technology of Greek and Roman Antiquity (1963) 204, cf. 146–7Google Scholar.

⁷³ A good example is the crane shown in the wellknown relief from the Tomb of the Haterii at Rome (Orlandos fig. 50). Taken literally, this shows a crane with five strands of rope hanging from the jib, and so five pulleys; it has a tread wheel with a diameter 12 times its axle diameter, and driven by 7 men. If we assume that each man is able to exert a force of 50 kg on the circumference of the wheel, the theoretical maximum load of the crane will be 5 × 12 × 7 × 50 kg = 21 tons. Although this would be substantially reduced by friction, it would be possible to have 6 pulleys instead of 5, a treadwheel diameter more than 12 times the axle diameter, and considerably more than 7 men effectively working the wheel, so that the real lifting capacity could still be of the order of 20 to 30 tons.

⁷⁴ The argument to be used below is not invalidated if the specific figures given in this paragraph are not accepted. It is sufficient to accept that there is some limit on the power input, and so the lifting capacity, of a winch and pulley hoist.

⁷⁵ It is interesting to speculate, but difficult to calculate, how far the expense of constructing a ramp exceeds the expense of building a heavy crane from scratch, given that no lifting device is already available on the site, and that there is at the time no intention of constructing more than one building. We usually assume that the ramp has to be specially built, while the crane is already there.

⁷⁶ An important point in the comparison is that the expense of making a ramp of given slope increases at a rate somewhere between the square and the cube of the increase in height. Increasing the height of a crane involves much less rise in the cost, but here again the limit feasible with simple means is reached sooner with a crane than with a ramp.

⁷⁷ See for example note 49.

⁷⁸ The lost relieving lintel over the central door of the Propylaia had a total volume of 6·3 m³ and so a weight of c. 17 tons, but like the other lintels of the Propylaia it probably consisted of two blocks side by side, so that each would weigh about 8½ tons.

⁷⁹ Clarke, S., Engelbach, R., Ancient Egyptian Masonry (1930) 148Google Scholar; cf. Engelbach, R., The Problem of the Obelisks (1923) 66–79Google Scholar.

⁸⁰ Balanos, N., Les Monuments de ľ Acropole, Relèvement et Conservation (1938) 85Google Scholar, Parthenon folding plate 1. Penrose, F. C., The Principles of Athenian Architecture (2nd ed., 1888) pl. 16Google Scholar wrongly shows the lintel formed of three blocks.

⁸¹ KP 121–7.

⁸² KP 154–66.

⁸³ The architraves may have been set no earlier than those of the Olympieion at Akragas, but the crucial stage was the decision to build a colossal temple on conventional lines, a decision which must have been taken in about 530–20 B.C., probably before cranes were exploited in architecture. Once a temple of this type was begun, there was no way of avoiding the huge architrave blocks, even though they demanded procedures which, by the time they were set in place, might seem old-fashioned.

⁸⁴ The lines indicating drum divisions in the restored elevation of the temple (KP 126, fig. 105) may not be absolutely accurate, but the photograph (ibid., fig. 107) suggests that they are roughly correct; in that case some of the drums used weighed over 50 tons.

⁸⁵ Penrose, F. C., The Principles of Athenian Architecture (2nd ed., 1888) 37–8Google Scholar. Penrose (ibid. 18) notes that the marble is coarser-grained than in Periklean buildings: but it is still no doubt Pentelic.

⁸⁶ Wiegand, T., Didyma 1 (1941) 98–9, pl. 63Google Scholar.

⁸⁷ Those of the Basilica Nova, listed here, seem to have been the heaviest at Rome, but the monolithic shafts of the Pantheon, the temple of Antoninus and Faustina, the temple of Saturn, the Baths of Caracalla and Diocletian all seem to have weighed between 30 and 50 tons.

⁸⁸ Wiegand, T., Baalbek 1 (1921) pl. 23Google Scholar.

⁸⁹ Ibid., fig. 30.

⁹⁰ The yard arms of Dionysos' boat on the cup by Exekias are each controlled by a rope passing from the stern through a loop at the yard arm and back to the stern again (Casson, L., Ships and Seamanship in the Ancient World (1971) 69–70Google Scholar). No pulley is shown at the yard arm, but the helmsman hauling on one end of the rope, the other being fixed, would gain some mechanical advantage. The principle is the same as in a compound pulley hoist, and once noticed, the same principle could easily have been used in loading and unloading cargo, with pulleys introduced to reduce friction. But there is no evidence whether that was in fact done in the sixth century B.C. By the first century B.C., of course, a block and tackle was certainly used for handling cargo (Vitruvius 10.2.10).

⁹¹ I am greatly indebted to Dr W. H. Plommer for his detailed and helpful criticisms of a draft of this paper. He has enabled me to tighten up the argument in many places, and the weaknesses that remain are there in spite of, rather than because of, his advice.

⁹² See note 78.