Ever since their discovery by Hormuzd Rassam in 1878 the Balawat Gates of Shalmaneser III (858–824 b.c.) have attracted the attention of archaeologists and art historians. The sixteen bronze bands that originally decorated a pair of massive wooden gates have embossed and chased decoration showing campaigns and the receipt of tribute from places including Tyre, Sidon, Hamath, Carchemish, Urartu, Babylonia and Bit-Adini on the River Euphrates. There are also expeditions to Lake Van or Lake Urmia and the source of the river Tigris. The decoration on these bronze bands was published in full by W. de Grey Birch and T.G. Pinches in 1902 and again by L.W. King in Reference King1915. There are also lengthy cuneiform inscriptions on the bronze sheathings covering the door edges. These give a version of the annals of Shalmaneser III (Grayson Reference Grayson1996: A.0.102.5). There is one matter, however, that has proved to be controversial, and that is the arrangement of the bands on the wooden door leaves. For reasons that are described below, the order of the bands was not noted at the time of excavation, and over the years a number of hypotheses have been put forward by various scholars. A probable order for the bands on the Shalmaneser gates was proposed by the authors in the definitive publication of the Ashurnasirpal Gates from Balawat (Curtis and Tallis Reference Curtis, Tallis, Barnett, Davies, Howard and Walker2008: 11–15),Footnote ¹ and here the opportunity is taken to present some new and we believe compelling evidence to support that arrangement, albeit with some modifications.

Rassam conducted excavations at Balawat on behalf of the British Museum between January and April 1878.Footnote ² He was drawn to the site by the recent discovery there of fragments of embossed bronze band with narrative scenes, found during the digging of a grave or graves. In fact, the mound of Balawat is covered with graves (Curtis Reference Curtis, Werr, Curtis, Martin, McMahon, Oates and Reade2002), and this was to cause great difficulties for Rassam. Nevertheless, in the course of his excavation his discoveries included the pair of gates set up by Shalmaneser III, the top bands of which had been found by the grave-diggers, a smaller set of gates set up by Ashurnasirpal II, a stone chest with three inscribed stone tablets in what we now know was a Temple of Mamu, and an Assyrian well. The main sources for the excavations are articles by Rassam in the Journal of the Transactions of the Victoria Institute (Reference Rassam 1881 ) and in the Transactions of the Society of Biblical Archaeology (Reference Rassam 1882 ) and in his book Asshur and the Land of Nimrod (Reference Rassam 1897 ).Footnote ³ In the latter, Rassam describes the discovery of the Shalmaneser Gates, which is what concerns us here, as follows (1897: 207–8):

The decorative bands were in due course published by Birch and Pinches (Reference Birch and Pinches1902), Billerbeck and Delitzsch (Reference Billerbeck and Delitzsch1908), Unger (Reference Unger1913; Reference Unger1920) and King (Reference King1915). From these studies it became clear that the two wooden gates were each about 7.92 metres in height and about 7.5 centimetres thick. The gates were fixed to huge wooden posts that revolved in stone sockets. At the back of each door there must have been wooden battens holding the planks of wood together. On each door-leaf there were eight bronze bands that ran across the flat part of the door and then around the post. At the end, each band was cut out to fit round the batten on the back of the door. The bronze bands have short inscriptions or epigraphs identifying the subject matter (Grayson Reference Grayson1996: A.0.102.63–86). The front edges of the door-leaves were covered with bronze sheathing on which were lengthy cuneiform inscriptions. Originally there would also have been bronze sheathings on the top and bottom edges of the door-leaves, but only those from the bottom now survive.Footnote ⁴

Unfortunately, when Rassam excavated the monument he made no record of the order in which the bronze bands had been fixed to the wooden doors. Obviously, the pieces of band that were found during the digging of graves and are now distributed between the Louvre Museum, the Walters Art Gallery, Baltimore, the Museum of the Ancient Orient, Istanbul, and the British Museum, must have been at the top of the gates and they would have belonged to Bands 1 and 2 on the left and right side of the gates. However, controversy has surrounded the arrangement of the other twelve bands, all of which are now in the British Museum. Birch and Pinches (Reference Birch and Pinches1902) proposed a chronological arrangement, but King (Reference King1915: 15, n.) was dismissive of this theory, writing “it is clear that the bands were not arranged upon the gates in chronological order. As they were designed for decorative effect, it is obvious that their arrangement would have been determined by the character of the reliefs upon them, with little, if any, reference to the relative order of the campaigns”. He believed that the less well finished bands such as his Band VIII (BM 124663) would have been fixed on the upper part of the doors, well above eye level; he suggests (1915:15) this hypothesis is supported by the fact that on the less well finished bands the figures are larger than those on the lower bands so that when seen from ground level they would appear to be of the same size as the figures in the lower bands. It is certainly true that King's Band VIII (our Band R2) almost certainly belonged at the top of the gates, but to some extent any judgment about the quality of the workmanship is bound to be subjective. Also, if it is really the case that the figures in the top, poorly finished bands are larger than the figures in the lower, better finished bands, it would imply that the bands were made to order following a carefully prescribed scheme, which seems doubtful.

The most plausible reconstruction was put forward by Eckhard Unger (Reference Unger1913; Reference Unger1920). He argued that the gate-posts to which the door-leaves were attached must have been made of gigantic cedar logs that would have tapered towards the top. Consequently, those parts of the bronze bands that circled the door-posts would have become progressively shorter from the bottom of the gates upwards. The crucial measurement for Unger was the distance from the bend where the band starts to go round the post (Knick) to the start of the U-shaped cut-out (Zungenspitz). These measurements would have decreased between the bottom and top of the gates. At the same time, if the posts were upright the lengths of the bands themselves would have increased from bottom to top. Unger then used these measurements to suggest an arrangement for the bands. He also suggested that the bands would have been arranged in four groups of four leaving a gap in the middle of the doors (1929: 99, fig. 5). Hans Güterbock (Reference Güterbock1957: pl. 22) followed Unger's order but thought the bands must have been spaced evenly down the doors rather than arranged in groups.Footnote ⁵

The theory on which Unger based his reconstruction is, we believe, essentially correct, but Unger was hampered by not being able to make a close inspection of the gates themselves. He had to base his measurements on photographs and on what he was able to see through the glass of the showcases. Inevitably Unger's measurements were in some cases approximations, so in the publication of the Ashurnasirpal Gates (Curtis and Tallis Reference Curtis, Tallis, Barnett, Davies, Howard and Walker2008: table 2.1) we used the opportunity to publish new and accurate measurements of the flat parts of the bands and those parts of the bands encircling the posts. In this way, we arrived at the arrangement presented in Table A.

Remarkably, this arrangement was almost exactly the same as Unger's, the only difference being that we transposed Unger's R1 and R2 and his R5 and R6.

During the preparation of the publication of the Ashurnasirpal II gates (Curtis and Tallis Reference Curtis, Tallis, Barnett, Davies, Howard and Walker2008) it became apparent that there was another way of establishing the order of the bands. As has been explained, after the bands had been mounted with nails on the wooden leaves, bronze edging strips were hammered on to the top, bottom and front edges of the doors (Fig. 1). These strips overlapped the edges and were folded over onto the flat surfaces of the doors. In this way, at the front edges of the doors the edging strips covered the ends of the bands. The edging strips were held in position by nails driven at regular intervals into the wooden doors, and wherever there were the ends of bands beneath the edging strips the nails were driven through them as well. As emphasized in the authors’ publication of the Ashurnasirpal II Gates, the gates are craft products, made with great skill but without precise measurements of every element. Thus, it was noticeable when studying these gates that the nails holding the edging strips in position were not placed in straight lines, actually far from it, and that every grouping of nail holes punched in each band end and its section of edging strip was, in fact, unique. It was even possible to match a photograph of a band end from the Temple Gates with a photograph of a fragment of its original, but detached and now dissociated section of edging strip merely from sight. Therefore, it occurred to the authors at this time that this could be a way of independently verifying the arrangement of the Shalmaneser bands in addition to Unger's scheme. In fact it was found that the Shalmaneser Gates were made with much greater care than the Ashurnasirpal II Gates, but even allowing for this, the pattern of holes for each band is unique and wherever possible an attempt was made to match the complete bands with their original locating holes in the edging strips.

Fig. 1 Electrotype copies of the Balawat Gates of Shalmaneser III mounted on modern wooden doors and ordered by their British Museum numbers. The edging strips are in the centre of the doors

Firstly, the edging-strips were carefully drawn. There are five sections (Fig. 2) preserved in the British Museum as follows:Footnote ¹⁹

1. 1. BM 124666/Rm 1046, text B v–vii. L. 191.8 cm (with further surviving part consisting of cuneiform inscription only 205.2 centimetres). From left-hand gate, near top (see Fig. 2).

2. 2. BM 124669/Rm 1049, text B iii–iv. L. 213.8 cm. From left-hand gate, middle (see Fig. 2). At the top of this edging-strip there are finished edges showing that the edging strip on the left was made of at least two pieces. On exhibition British Museum July 2015.

3. 3. BM 124665/Rm 1045, text B i–iii. L. 165.1 centimetres. From left-hand gate, bottom (see Fig. 2). This piece is mounted on a timber beam together with a further section of edging strip (L. 70.4 centimetres) represented only by the front inscribed surface. As mounted, the complete piece is separated from the inscribed piece by a gap of 42.0 centimetres. The length of this gap has presumably been estimated on the basis of the amount of cuneiform inscription missing. The complete length of pieces mounted on beam no. 3 is therefore 277.5 centimetres. There are finished edges at the bottom of the strip showing that it is from the bottom of the left-hand door. There is a nail-hole at the bottom for the sheathing that would have been fitted along the bottom of the door.

4. 4. BM 124668/Rm1048, text A ii–iii. L. 221.0 centimetres. From right-hand gate, near top (see Fig. 2). There are finished edges at the bottom of this edging strip which would have occurred about half way down the gate showing that the edging strip on the right was also made of at least two pieces. On exhibition British Museum July 2015.

5. 5. BM 124667/Rm 1047, text A iv–vi. L. 318.2 centimetres. From right-hand gate, middle and bottom (see Fig. 2). It was possible to remove part of the sheathing from the timber beam on which it was mounted and it could be seen that on the underside of the top surface (i.e. on the underneath of the surface with the rosettes) there were shadows in the corrosion in positions R5 and R6 showing where the ends of the bands had been under the edging strip. In these areas the surface is grey-green in colour, as opposed to yellow-green elsewhere, quite smooth and much less corroded than the surrounding areas. In one case (R6) the corrosion shadow was complete and was c.28 centimetres wide, while in the other case (R5) it was only partially preserved, covering two holes. It was not entirely clear whether the differential corrosion extends right up to the bend in the metal. Two further pieces of bronze edging strip were originally mounted on the same timber beam, but they are now missing.Footnote ²⁰ These pieces have lengths of c.46 centimetres (at the top) and c.31 centimetres (at the bottom); they have not been included in the drawing, but would give a total length for this piece of c.395 centimetres. There are two places in the surviving length of strip where the metal abuts, or almost; these are not physical joins, but the positioning is presumably confirmed by the cuneiform text.

Fig. 2 Drawings of bronze edging-strips showing their relative positions and where the bands were originally located. 1. BM 124666/Rm 1046, text Bv–vii; 2. BM 124669/Rm 1049; 3. BM 124665/Rm 1045, text Bi–iii; 4. BM 124668/Rm1048, text Aii–iii; 5. BM 124667/Rm 1047, text Aiv–vi.

The edging-strips are folded around the front edges of the doors to a depth of about 6.0–6.5 centimetres at front and back and with a width of about 8.9 centimetres at the front. They are held in position by nails at the front and the back. There are long cuneiform inscriptions on the front edges of each door leaf. They duplicate each other, but that on the left runs from the bottom upwards, and that on the right runs from the top downwards. These cuneiform inscriptions contain the annals of Shalmaneser III up to 850 b.c. (Grayson Reference Grayson1996: A.0.102.5).Footnote ²¹ At the back, the edging-strips are indented at regular intervals (the “cut-outs”) to accommodate the battens that are fixed to the backs of the doors and hold them together (Fig. 3). These battens were held in position by nails probably driven from the front and back of the doors, so on the front of the doors there would probably have been horizontal rows of nail-holes at regular intervals for the whole height of the doors. The cut-outs in the edging-strips measure c.12.0–12.5 centimetres × c.2.5–2.8 centimetres and are at intervals of c.32–35 centimetres. It was established from corrosion shadows on the underside of edging-strip no. 5 in positions R5 and R6 (Fig. 4) that the bands on the front corresponded to every other cut-out on the back; in other words, the bronze bands would probably have covered alternate rows of nail-holes on the front of the gates. In this way, it was possible to work out where all the bands would have been placed (but not their order) in relation to the edging-strips (see Fig. 2). The bands, generally about 26 centimetres in height, would then have been about 65 centimetres apart. On the front of the doors the edging-strips were held in position by nails driven through the centre of embossed rosettes arranged in a vertical line (but not always exactly aligned) for the whole height of the door (Figs. 5–6). These rosettes measure c.2.6–2.8 centimetres in diameter, and the central holes are at intervals of c.4.5–6.0 centimetres. The rosettes are made with a punch, with each petal having been punched out separately. There is thus some variation in the number of petals: usually there are about thirteen to fourteen but there can be as many as twenty. The variation in the form and number of petals shows that there were a number of craftsmen at work. The petals surround an incised circle c.1.5 centimetres in diameter. The holes in the centre of the rosettes are generally c.0.5 centimetres in diameter.Footnote ²² In all cases the edging-strips overlaid the ends of the bands, so wherever there were bronze bands under the edging-strip the nails went through the bands as well; there are generally five nail-holes in the end of each band.

Fig. 3 Section of edging-strip BM 124669 showing cut-outs and cuneiform inscription

Fig. 4 Corrosion shadow on the underside of the edging-strip in position R6 showing position of a band, probably BM 124662

Fig. 5 Section of edging-strip BM 124669 showing rosettes with nail-holes and cuneiform inscription

Fig. 6 Close-up view of rosette with nail-hole (BM 124669)

In some cases the decorated ends of the band are covered up and holes made through the decorated parts. For example, on our Band L6 (BM 124656, King Reference King1915: pl. LIX), showing the expedition to the source of the Tigris, the soldier in the lower register would have been completely covered up, and on our Band L3 (BM 124651, King Reference King1915: pl. XXX) showing tribute from Unqu, part of a camp and part of a city would both have been covered up. In other cases the decoration ends before the edging strip (e.g. our Band R6, BM 124662, King Reference King1915: pl. I) and in some cases there is even a substantial margin (our Band R7, BM 124657, King Reference King1915: pl. LXXII). Where the embossed linear borders go under the edging strip they have sometimes been flattened at the ends (e.g. our Band L8, BM 124654, King Reference King1915: pl. LXXI), either through the action of hammering down the edging strip or perhaps in anticipation of their being covered up by the edging strip. In several cases there is the impression of a vertical line on the band clearly showing the original position of the edging strip (e.g. our Band R7, BM 124657, King Reference King1915: pl. LXXII).

Having established the probable location of bands, those sections of edging-strip where the bands were thought to have joined were carefully redrawn and the arrangement of the holes was compared with the holes in the band-ends (see Figs. 7–8).

Fig. 7 Drawings of sections of edging-strip from left-hand door with ends of bands showing position of nail-holes

Fig. 8 Drawings of sections of edging-strip from right-hand door with ends of bands showing position of nail-holes

On the left-hand door-leaf, the edging strip is incomplete at the top so that part around position L1 is completely missing. The strip is partially preserved around position L2 but not well enough for it to be matched with a band-end. In any case, the two bands at the top of the gate on the left side are both fragmentary. The other 6 bands on the left (L3–8) are all well-preserved at the right-hand end where they would have been under the edging-strip, but the edging strip is completely missing in position L6, and partially missing in positions L3 and L7. The possibility of a complete match was therefore only available for positions L4, L5 and L8, but we also attempted a match, albeit partial, in positions L3 and L7 (see also Fig. 9). In all cases there was a perfect alignment, which confirmed the previously suggested arrangement:Footnote ²³

On the right-hand door-leaf, the edging strip is again missing at the top so we do not have that section where Bands 1–2 would have fitted. Of the other six bands, BM 124659 (previously R5) is not preserved at the left-hand end where it would have fitted under the edging-strip and the edging-strip is partially missing/in bad condition in position R8. It was discovered that bands BM 124661 and BM 124658 were excellent matches for positions R3 and R4, as previously proposed, but beyond this there were problems. Although the edging strip is damaged at position R8 it was immediately apparent that BM 124657 could not fit here as previously believed. Instead, BM 124657 was found to be a good match for position R7. This was confirmed not just by the drawings but also by a Photoshop overlay (Fig. 10). Following this discovery, BM 124662 was aligned with position R6 and BM 124660 was aligned with position R5. The band missing its left-hand end, BM 124659, that had previously been assigned to position R5, was then relocated to position R8; this was still in keeping with the general principle of the Unger scheme because its band around the post measurement is greater than 81.7 centimetres, so it could have been greater than the figure given for the new Band R7 which is 87.2 centimetres (see table 2.1 in Curtis and Tallis Reference Grayson2008). The results of these changes, then, is that the bands that were previously thought to be R6, R7 and R8 each move up one, and the band previously thought to be R5 becomes R8.Footnote ²⁴ The new arrangement is therefore as follows:

While the principle of checking the order of the bands by matching the nail holes at the ends of the bands with the holes in the edging strips is undoubtedly sound, and we are fairly confident that the order now proposed is the correct one, it has to be acknowledged that the conditions for the enquiry were not perfect. Firstly, the bronze edging strips are now mounted on long pieces of wood so that the underside of the bronze sheathing cannot be seen. Ideally, the bronze edging strips would have been taken off the wooden supports to inspect and record the shadows in the corrosion on the underside where the band ends fitted under the edging strips, but in most cases this was not practical. Consequently, we were only able to record the different patterns in the corrosion caused by the band-ends in two cases (positions R5 and R6). Secondly, we did not have the opportunity to make exact 1:1 drawings of the band ends from the gates themselves, which could then have been compared with the exact 1:1 drawings of the edging strips. Instead we made drawings from the excellent photographs published in King Reference King1915 and then scaled them up until they fitted the relevant row of holes in the edging strip. In this way for the bands drawn we arrived at band heights of 25.7–28.5 centimetres, with an average of 26.9 centimetres.Footnote ²⁵ This contrasts with figures of 24.8–27.2 centimetres for Bands 3–8 on each leaf (given in Curtis and Tallis Reference Curtis, Tallis, Barnett, Davies, Howard and Walker2008: table 2.1 on p. 14), with an average of 26.7 centimetres. In the circumstances such a small discrepancy seems quite acceptable, particularly as in all cases the height of a band varies along its length and it is not always clear where the measurement was taken.

The proposed changes in the arrangement of the bands do of course have important implications for the overall decorative scheme on the gates. We now have an arrangement as follows ( table B and Fig. 11):

Fig. 11 Photographs of the bronze bands from the Shalmaneser III Gates ordered according to the newly suggested arrangement

The pairings for Bands L1–R1, L2–R2, L3–R3 and L4–R4 are unchanged and are as recorded in Curtis and Tallis Reference Grayson2008: table 2.1 on p. 14, but those for Bands L5–R5, L6–R6, L7–R7 and L8–R8 are now all different.

What can be deduced from this arrangement? Firstly, as is clear from the table above the bands are clearly not arranged in a chronological order. In fact from a chronological point of view they seem to be random. For example, campaigns between 858 and 854 b.c. (Bands R2–R4) are bracketed between bands showing a campaign or campaigns in Babylonia in 850 b.c. (Bands R1, R5). So is there a geographical arrangement? This does not seem to work either. Thus, we have campaigns in Babylonia on Bands R1 and R5, and campaigns in Hamath on Bands L1, L5, and R7. It is true that there does seem to be a preponderance of campaigns in the west in the upper part of the gates (Tyre and Sidon on Bands L2 and R3, Syria on Bands L1, L3, L4, R4 and L5), and a preponderance of campaigns in the north in the lower part of the gates (source of Tigris on Band L6, Urartu on Bands R6, L7, R8), but this can only have been a general principle and was not rigidly adhered to. So, we have Shubria (located north of the Upper TigrisFootnote ³⁹ ) near the top of the gates and Bit-Agusi (in North Syria) at the bottom of the gates. Surprisingly, the bands are not even arranged in pairs. For example, one might have expected bands showing Tyre and Sidon and bands showing Urartu to be opposite each other but this is not the case. The only pairings that could perhaps have any significance are Bands L4 and R4 (Carchemish and Bit-Adini) and Bands L6 and R6 (sacrifices at the source of the Tigris and the Sea of Nairi).

Given that the gates do not seem to be arranged chronologically or geographically (at least, certainly not rigorously), is there some other guiding principle? It is interesting—it cannot really be put more strongly than that—that in some cases a vaguely symmetrical arrangement can be noted. Thus, both Bands L1 and R1 show scenes of warfare, and Bands L3 and R3 both show towns surrounded by water and the delivery of tribute. Bands L4 and R4 both show camps in the lower register, one square and one round, with Assyrian soldiers proceeding from them. Bands L5 and R5 have square camps in the lower register facing each other, again with Assyrians proceeding from them. In the upper register on the left there are charging horsemen and chariots and on the right there are horsemen on the march. Bands L6 and R6 show expeditions to the source of the Tigris and the Sea of Nairi respectively, with sacrifices being offered in both cases. Lastly, the four lowest bands (L7, R7, L8, R8) all show scenes of warfare with the king receiving booty and prisoners. However, it is not really possible to put forward an argument for a symmetrical arrangement any more than for a chronological or a geographical arrangement, so we are obliged to conclude that if there was a rationale behind the arrangement of the bands it eludes us at present.

The above article represents an attempt to draw attention without further delay to a method for determining the arrangement of the bronze bands on the Shalmaneser III gates. This does not replace the methodology devised by Eckhard Unger, with which we are generally in agreement, but it is an alternative technique for working out where the bands should be placed. Broadly speaking, this new technique validates and confirms Unger's conclusions. We are fairly confident that the newly proposed arrangement (table B) is correct, but there may be still be scope for some minor adjustments. It has not been the intention in this short article to consider in depth the style or content of the bands, and for these matters the reader is referred to the important studies of Schachner (Reference Schachner2007), Hertel (Reference Hertel and Dercksen2004), Reade (Reference Reade1979: 64–72), Marcus (Reference Marcus1987), Wicke (Reference Wicke2012) and others.