Ancient History & Civilisation

Appendix 3

Building the Ark – Technical Report

(with Mark Wilson)

To safeguard the world’s largest boat

They smoothed on a bitumen coat

They brought in the oracle

Who said, of their coracle,

Though dry I doubt it will float.

C. M. Patience

Atra-hasīs’s Ark

The following notes on the text of the Ark Tablet look at each building section in turn, supporting what can be gleaned from the tablet and interpolating from construction accounts of other similar traditional vessels. For clarity, the calculations involved are carried out in Babylonian units. We take ‘one finger’ as our basic unit of length, after the Babylonian ubānu ‘finger’ measure which is used in the Ark Tablet. One Babylonian finger is approximately 12/3 cm and it is usual to take it as exactly that for ease of calculation.



1 ubānu = 1 finger ≡ 1.666 cm

1 ammatu (cubit) = 30 fingers

1 nindanu = 12 ammatu = 360 fingers


1 ikû = 100 (= 10 × 10) nindanu2 = 12,960,000 fingers2 = 3,600 m2


1 qa = 216 (= 6 × 6 × 6) fingers3 = 1 litre

10 qa = 1 sūtu = 2,160 fingers3

1 gur = 300 qa = 64,800 fingers3

1 šar = 3,600 (sūtu) = 7,776,000 fingers3

‘Floor area’ is qaqqaru, ‘ground’, which also has the more specific meaning ‘surface’, or ‘area’. Here it means the floor of the boat, as we are told in the technical dictionary:

Sumerian giš-ki-má = Babylonian qaq-qar eleppi (GIŠ-MÁ), ‘wooden floor of a boat’

crib: giš = īṣu, ‘wood’; ki = qaqqaru, ground’; má = eleppu, ‘boat’


The fundamental facts regarding the Ark are given in lines 6–9. The Ark has a circular design, and is to be built inside a circle drawn out on the ground. We are told that its base area is one ikû, and that its walls are one nindan high. That is (using Area = π × Radius2), its diameter is 67.7 metres, its walls six metres tall. As it is essentially a giant coracle, its construction methods have been compared with those of the traditional Iraqi coracle, or guffa, as reported by Hornell.

This record-breaking guffa differs from its conventional relatives in several ways. Chief among these is the existence of a roof, obviously indispensable. Although the roof is not explicitly mentioned in the construction details, we are assured of its final presence by the fact that we are told that Atra-hasīs goes on to it to pray later on in the tablet.


Lines 10–12 give the materials for the hull of the boat, and these are described as the ‘ropes and rushes of a boat’.

The ropes: kannu

This word, kannu ‘A’, means a fetter, band, rope, belt or even a wisp of straw. It can be tough enough to restrain a runaway slave or make a wrestler’s belt, and slim enough to be a hair band. The verbal root from which it derives, kanānu, means ‘to twist’, or ‘to coil’, which is natural for a word meaning ‘rope’.

The rushes: ašlu

There are two identical-looking words pronounced ašlu. ‘A’ means ‘rope’, ‘tow rope’, ‘surveyor’s measuring rope’; ‘B’ means a kind of rush which can be used to make matting for furniture but also, in narrow quantities, as a thread or twine. This is the ašlu we want. It is written with a complex cuneiform sign that also serves for other types of rush, and is thus distinct from ašlu A, which is a true rope.

The structure is thus made totally of plaited palm-fibre rope and rushes, whose intrinsic twisting and interlacing immediately suggests the process of basket weaving, and from this we conclude the form to be a giant coiled rope basket. That this is produced before any internal framework is consistent with Hornell’s account of how a traditional Iraqi guffa is manufactured.

As well as the material for the basket – palm fibre – we are also told its required volume. This rope volume is 4 šar (4 × 3,600) + 30 = 14,430 ‘units’ of material to make the basket alone.

We defend here our conclusion that the quantities written in šār are meant to be and have to be taken seriously. A thousand years after the Ark Tablet was written these numbers in Gilgamesh XI became fanciful expressions to convey great magnitude, although an intervening scribe might have back-calculated certain measures in the attempt to reconcile textual variants. The important issue – which crops up several times in the Ark Tablet text – is that quantities of a raw material are given solely as totals with the standard units only implied. Here we have found that out of the two likely choices – sūtu or gur – for the standard volume measure behind numbers in šār, only sūtu gives meaningful results.

Given the shape and size of the Ark’s basketwork hull, we substantiate this claim about the accuracy and nature of the numbers by comparing the amount of material that should be used in such a construction – which we call VCALCULATED – with the amount given in the text – which we call VGIVEN

To perform this calculation we need two additional items of information. The first is the thickness of the woven rope basketry. Although this is not given in the cuneiform text, a clue comes from the (partly restored) ‘ropes … for [a boat]’ (line 10), implying, very plausibly, a type of rope peculiar to boat-making whose thickness was no doubt standard. The text also tells us the ropes were to be made by someone other than the shipwright, presumably a ‘boat-rope’ artisan, who would manufacture rope of a type independent of the size of coracle it was for. We assume that the thickness of rope used to make a guffa was always independent of its size, so that the width of the rope ordered would not scale up with the final size of the boat. Indeed, the descriptions of traditional guffas show that their structural rigidity depends on the internal framework, so that the basket is just a skin of convenient material to support the applied waterproofing layer. Assyrian sculptures mentioned in Chapter 6 prefer skins to rope for the hulls of their guffas.

This means the basket body of the Ark is constructed using standard materials and techniques, and, although it is nearly seventy metres across, the walls are still seen as having the same thickness as conventional-sized coracles. The most likely standard width of rope used is one finger, which is supported by early photographs of Iraqi guffas (e.g. ‘Building the peculiar round boat …’), which show that the rope used was about one finger in thickness. This is supported by another calculation below about bitumen.

A new Iraqi coracle has just been finished. The rope’s thickness can be seen to be roughly the thickness of a finger (or a toe). High-quality stereo photographs of this type from the 1920s preserve important information which often is otherwise unobtainable. Close-up of modern reed basket-work.

(picture acknowledgement app.1)

The second piece of information we need is the cross-sectional curve of the walls. These should have an outward camber at the base to resist hydrostatic pressure, and this is what is seen on the photographs of actual guffas. There the curvature of the walls is seen to lie somewhere between a straight-sided cylinder and the semicircle of the outer half of a torus (ring-doughnut). Therefore, we believe that it would not be far off the mark to assume it was exactly halfway in-between, and approximate the curvature by a semi-ellipse whose width is a quarter of its height. This means the walls of our reconstructed Ark – the sides of which are one nindan high – bulge out from the base by one-quarter of a nindan at their maximum diameter, thus:

As is approximately true of real guffas, the walls are symmetrical about the mid-transverse plane, meaning the Ark would look the same if it was turned over, top-to-bottom. The important corollary to this is that the area of the roof is identical to the area of the base.

Rope calculations

The first step for rope volume used is to calculate the total surface area ‘A’ of the vessel. This is area of the base ‘B’, plus the area of the roof ‘R’, plus the area of the walls ‘W’.

We are given B = 12,960,000 fingers2, and have assumed that R = B. To calculate the area of the walls W we need Pappas’s First Centroid Theorem: The surface area W of a surface of revolution generated by rotating a plane curve about an axis external to it and in the same plane is equal to the product of the arc length L of the curve and the distance D travelled by its centroid (centre of gravity): W = L × D.

Here, the plane curve is the semi-elliptical shape of the walls, and its length is just half of the circumference of the full ellipse of which it is a part. Calculating the circumference of a general ellipse is a nightmare of complexity, but for the specific case we use here – one whose major axis of length ‘a’ is twice as long as its minor axis – we have a neat formula to use called Ramanujan’s Approximation which is correct to three places of decimals:

Ramanujan’s Approximation

Here, a is simply the height of the walls, 360 fingers, and we are only interested in half of the circumference, which gives us L ≈ ½ × 2.422 × 360 = 436 fingers.

The other component we now need is D, the length travelled by the centroid as the semi-ellipse is rotated to form the walls of the Ark. This is the length of a circle swept out by a radius equal to that of the base of the Ark plus the additional distance from the edge of the base to the centroid. We know the base of the Ark is a circle of area one ikû, so (from ‘Area = π × Radius2’) we can calculate its radius ‘r’ to be:

r = √(Base Area B/π) = √(12,960,000/π) ≈ 2,031 fingers (working to the nearest whole finger.)

The distance ‘d’ of the centroid of a semi-elliptical arc to the axis of the ellipse is given by:

Using the familiar rule for circles ‘Circumference = 2π × Radius’, we are now in a position to calculate D, the circumference of the circle travelled by the centroid:

D = 2π × (r + d) = 2π × 2,088 fingers ≈ 13,119 fingers.

Finally, this gives the area of the walls W as:

W = L × D = 436 fingers × 13,119 fingers ≈ 5,719,880 fingers2;

giving the total area of the Ark as:

A = B + R + W = 12,960,000 + 12,960,000 + 5,719,880 ≈ 31,639,880 fingers2

We now assume that the ropes are whipped tightly enough to each other that they are densely packed and their cross-section can be taken as square with negligible error. Similarly, since the basket is very thin compared to its area, we can calculate its volume by just multiplying its area by its thickness of one finger, again with negligible error.

Thus our calculated volume (VCALCULATED) of rope needed to make the basketwork of the Ark is:

VCALCULATED = 1 finger (thickness) × 31,639,880 fingers2 = 31,639,8800 fingers3 or, dividing by 2,160 to give units of sūtu:

VCALCULATED = 14,648 sūtu.

The given volume (VGIVEN) of rope according to Enki is:

VGIVEN = 14,430 sūtu

which differs from our calculated figure by just a little under 1½ per cent. This is a striking result, and we take it as evidence to support assumption that the quantities given in the Ark Tablet are factual.

We can work out the length of rope represented by VCALCULATED by dividing it by the assumed cross-sectional area of the rope:

Length of Rope = 31,639,880 fingers3/1 finger2 = 31,639,880 fingers = 527 km.

As pointed out earlier, this is roughly the distance from London to Edinburgh!

The Babylonian reckoning

The very closeness of the figures VCALCULATED and VGIVEN leads one to question how the Babylonians might have made their calculation of the quantity needed.

We believe the answer lies in the fact that one ikû is defined as an area equivalent to that of a square of ten nindan × ten nindan, thus making it easy to visualise the area in terms of such a square. This proposition seems to us reinforced by Enki’s actually saying:

Draw out the boat that you will make

On a circular plan;

Let her length and breadth be equal,

especially given the circle-in-its-square school diagram illustrated on this page above.

The Babylonians found it difficult to do accurate arithmetic involving circular measures due to their imprecise value for π. If we assume that for the sake of ease of calculation they visualised the one-ikû base of the Ark as a square, then the walls will now be four panels, each ten nindan long by one nindan high, and this would be topped off by a square roof identical to the base. A trivial calculation of the area of this shallow biscuit-tin shape allows us to give the volume of material needed to make it by multiplying it by onefinger thickness, as is done for the Ark above. If we call the volume ‘VSQUARE’ we find:

VSQUARE = 14,400 sūtu.

This is four šār exactly, a difference of 0.2 per cent from VGIVEN!

When first encountered, the ‘plus 30’ in the figure VGIVEN seems like an insignificant if not inexplicable quantity, but the above calculation underscores its critical importance, for without it, it could be argued that the intention was to make a square-based vessel, but the extra thirty sūtu shows this cannot be the case. However, the ‘square-based’ method was almost certainly how the Babylonian scribes ‘back engineered’ their figure for the volume involved given the shape. We can see this by doing the calculation for the volume of fibre needed for a circular-based vessel with straight vertical sides – a cylinder. As a circle has the smallest circumference which encloses a given area, the length of these walls will be less than the ‘square based’ value, resulting in an overall volume smaller than VSQUARE by about 2 per cent. As we saw from our figure for VCALCULATED, the extra area provided by the bulge in the walls slightly overcompensates for this 2 per cent, and empirical knowledge of this may have led the Babylonians to formulate a rule of thumb for such volume calculations of the type, ‘Calculate the volume for a square-based vessel then add an extra bit on’

The ‘extra bit’ is what we believe the role of the thirty sūtu in VGIVEN’s ‘4 šār + 30’ to be. Whether or not such a procedure as this was actually used by the ancient Mesopotamian shipwrights, it is easy to see how it would have been useful in the typical scribal tasks of calculating the amount of rope needed to manufacture a particular size of vessel, as well as the quantity of bitumen needed to waterproof it.

The obvious question which then follows is how did they arrive at a number for that ‘extra bit’? For the Ark this figure is ‘30 sūtu’, so a natural assumption is that this is thirty times some real amount used for regular guffas. One way of pursuing this idea is to apply the above techniques to a guffa whose diameter is thirty times smaller than that of the Ark.

The diameter of this craft would then be:

4,062 fingers/30 = 135.4 fingers,

that is, a little over two metres. The walls of the Ark would not scale down in the same way, as their height is determined by practicality, as must have been true for the different sizes of guffas. The ‘square-based’ version of this would obviously have walls 10nindanu/30 = 120 fingers long. We can now check what height of wall would give a difference (extra bit) of one sūtu between the round guffa and its square-based approximation, and see if this would be a practical size for this boat.

A slightly more involved calculation shows this height to be 34.4 fingers, about 58 centimetres. That is, this mini-Ark would have a diameter about four times the height of its walls, a proportion which seems reasonably safe and practical for a boat ferrying goods and people in calm water. Indeed, the photographs of traditional guffas being built show boats with very similar dimensions.

Given the simplicity of Enki’s exhortation to build a boat ‘as big as a field’, it seems unlikely that this measurement is seen as being a regular guffa scaled up by a factor of 900 (= 302). However, this is possibly how the figures were arrived at in the scribal exegesis of the story. It is known that boats of the period came in standard sizes thought to be related to their cargo capacity, and it may have been either noticed or calculated that some measures for the Ark could be derived from those of a standard boat of one-thirtieth the diameter of the Ark.


In parallel with the description of building a traditional guffa given in Hornell, the next stage of construction comes in where the main structural framework is fitted (lines 13 and 14). These are called ribs on the Ark Tablet, and are simply described as being ‘set in’, with no clue as to the exact process or their arrangement, or even of the material from which they are made.

The only hard rib information concerns dimensions: the length is given as ten nindan (sixty metres), while they are ‘as thick as a parsiktu-vessel’. A parsiktu is a volume unit equal to sixty qa, deriving from the name of the wooden vessel used to measure out grain in approximately sixty qa amounts. That thickness is meant here rules out understanding parsiktu in this instance in its common meaning as a volume unit. It must refer to the seldom-mentioned measuring vessel itself. As explained, we take its usage in this context as hyperbole corresponding to our ‘as thick as a barrel’, designed to be an awe-inspiring superlative showing on the spot how much bigger the ribs of the Ark are when compared to those of a normal-sized vessel. Clearly we are meant to understand some approximate size from this statement, so the obvious question here is ‘how thick is a barrel?’

A traditional square grain measure from Japan. Most such objects seem to be round.

(picture acknowledgement app.5)

Traditional grain measures come in a variety of sizes and shapes, the most common being a squat cylinder whose width is about the same as its height. If as a working model we take this to be the shape of a parsiktu with an interior volume of 60 qa and stout walls 2fingers thick, then the width across its mouth would be about 29.5 fingers, or 49 cm. However, given the lack of evidence for cooperage in Old Babylonian times, it seems much more likely that the shape of vessel used as a grain measure would be a simple box shape, like that shown in the above picture.

Only one known cuneiform text actually quotes the size of a parsiktu-vessel, and then only hypothetically. Significantly for the composition of the Ark Tablet, this is a school tablet with a problem in which the schoolchild has to calculate the depth of a 60-qaparsiktu-vessel which is four unspecified units ‘across’. Since they don’t mention ‘sides’ as they usually do in such problems, this is is likely to be a square-topped box, with the ‘across’ being the diagonal from corner to opposite corner. The units can really only be ‘stacked hands’ of ten fingers. Of course the problem does not take into account the thickness of the walls of a real measuring box, but if we again estimate this to be two fingers then an elementary calculation (40/√2) tells us its width along each side is 32.3 fingers, or 54 centimetres (and, solving the schoolboy problem, 18.2 fingers deep if you include the assumed thickness of the walls).

The ‘60 qa’ parsiktu-vessel reconstructed from a school problem text.

This is not so far from the quoted estimate for a cylindrical measuring vessel, and means we can take ‘as thick as a parsiktu’ to mean roughly one cubit (∼fifty centimetres) thick no matter what the parsiktu’s shape. The fact that the ribs were not described as onecubit thick indicates the use of the term parsiktu as an informal and easy-to-grasp literary device rather than an exact measure. The ribs of the Ark are thus ten nindan long and about thirty fingers wide.

As to their cross-sectional shape the cuneiform text is silent, but this is no doubt implicit in the name ‘rib’, which must have had a technical usage in boat-building. We can work out all we need to know from the corresponding elements in the traditional guffa, described by Hornell as ‘lathes’, meaning they have a thin rectangular cross-section. They are made of a resilient wood, and are sown into the basketwork of the hull under tension as the main source of rigidity in this structure. They run from the gunwales down the walls and across the base of the boat, but they are not all directed at the centre. Instead, each one of a series is offset from the angle of the wall so that they run parallel across the base to one side of the centre. These ribs are then interwoven with a second series set at 90o to the first, like so:

Plan view of the Ark with two series of ribs set in at 90o.

As more of these pairs of series at 90o are set in around the circumference, they not only strengthen the walls but build up a floor structure as well, which is later reinforced by pouring bitumen between the ribs. The scheme above uses six ribs from our total of thirty, so another four such sets need to be laid in, each rotated around the circumference by 360o/5 = 72o with respect to each other. Hornell tells us that the number used on the largest of the traditional guffas was twelve to sixteen, so the Ark uses about twice as many.

The curved walls have been shown above to be about 436 fingers long from top to bottom, so each 10-nindan-long rib will run down the wall and then approximately 8½ nindan along the base of the boat. The gap between ribs at the wall will be a quite large one of seven metres or so.

As these ribs in a normal-size guffa are thin springy strips of wood, the implication is that giant ones here are also intended to be wooden. Although there are no trees from the Ancient Near East of sufficient size for these to be carved from one piece, plank-sized sections can be scarf-jointed together, and, if the resulting ribs also had a shallow enough depth, they would be sufficiently flexible to interlace like the regular lathes. Given the comparative fragility of the basket walls, however, it seems improbable that such ribs could be fitted without damaging the hull unless they were pre-shaped into long, laid-back ‘J’ shapes.

Importantly, unlike the thickness of the shell of the boat (and, as we shall see later, its waterproofing) – where no concession was made to its exaggerated size – these structural elements do scale up in comparison to those of a regular guffa, in both size and number. The practical aspects of handling such huge structures seem to have had no interest to the authors, and no information is given as to how or with what they were to be installed into the hull.

In his description of guffa-building, Hornell tells us that between these main ribs shorter upright lathes the height of the walls are sewn to the inside of the basket to provide additional rigidity. These elements are not explicit in our description, but perhaps this absence is explained through the following step.


At this stage there are no supports for the roof of the coracle basket, which must be assumed to have been woven along with the rest of the boat. The next lines of the Ark Tablet address this in typically succinct fashion, describing the installation of a vast number of stanchions as support for an interior floor and the fitting of wooden cabins so that the occupants had upper and lower decks. The presence of more than one deck is the second way in which the Ark differs from simply being a scaled-up guffa.

The supports are half a nindan long and – in parallel to the previous line about the ribs – ‘half (a parsiktu) thick’, and they are described as being ‘made firm’ within the boat (lines 15–16). If for simplicity we assume them to have a square cross-section then this would have an area of about 15 fingers × 15 fingers = 225 fingers2 each. Although the greatest dimension of these elements is described as a length, their vertical nature is inescapable through the use of the term ‘stanchion’ (imdu, from the verb ‘to stand’). Other uses of this term cited in the Chicago Assyrian Dictionary I/J assure us that these supports are intended to be made of wood. The Ark Tablet tells us that one šār, i.e. 3,600, are to be installed. Although this sounds more like an arbitrarily large number, it turns out that this number would actually take up only a little over 6 per cent of the one-ikû floor space of the Ark, which is similar to the proportion of any building’s floor space taken up by supporting walls. Indeed, if this number is intended to be anything other than a literary device (as seems probable to us), these supports must have been thought of as having a placement designed to bear the load of the structures on the upper deck, rather than simply being arrayed across the floor like a forest.

Although this upper floor or deck is not mentioned as such in the text, we are assured that this is the purpose of the supports by their height – which is half the height of the Ark, by their shape, and by their number, which would be adequate for the purpose. We are next told that cabins have been constructed ‘above and below’, and it is possible that the flooring of the upper cabins was simply meant to be understood as the upper deck, resulting in an economy of description. This floor level would bisect the internal space of the craft into two roomy decks each about three metres high.

Such boat cabins are usually described as being wooden, but this probably meant having a wooden framework with woven basketwork walls, an idea reinforced by the root of the verb used for their construction – rakāsu – which involves the idea of ‘tying’. The cabins complete the structural elements of the Ark, and result in a cross-section for the vessel which may be schematised thus:

The Ark showing its stanchions, deck, and upper and lower cabins.

Obviously the framework of the cabins on the upper deck will have the function of supporting the already-completed roof of the Ark. If the internal floor was extended until it could be fixed to the external walls, this would also increase the structural strength and more than make up for the absence of the shorter supports expected between the ribs. So the presence of a deck and roof results in a more robust craft.

Caulking the Ark

The next step toward the completion of the boat is the waterproofing of both outside and inside faces of all external walls. This is done with the two types of bitumen – iṭṭû-bitumen and kupru-bitumen – with a final coating of oil. Before we move on to what the tablet says about this procedure, it will be of benefit to look at what is known generally about the two bitumen agents.

There are two useful resources here. The first is Leemans 1960, which looks at tablets dealing with the waterproofing of boats, and tentatively deduces the following information, valid for the Old Babylonian period to which the Ark Tablet dates:

1. iṭṭû-bitumen was moist; kupru-bitumen was harder and more mastic;

2. iṭṭû-bitumen is used as a liquid for some tasks, and its liquid form is produced in a kiln;

3. for caulking boats, large amounts of kupru-bitumen are used in comparison with iṭṭû-bitumen;

4. For caulking, iṭṭû-bitumen could be used on a rough kupru-bitumen base to improve its quality;

5. iṭṭû-bitumen was used on top of kupru-bitumen, on cabins and on the inside.

The second source is Carter 2012, where analysis shows that ancient bitumen samples used in caulking were never just pure bitumen but included organic and mineral components in amounts suggesting they had been deliberately added, perhaps as tempering agents. In addition, quite large quantities of oil are accounted for in boat-building but for an unknown use, although it is assumed waterproofing of ropes might be involved.

Now we turn to what the Ark Tablet has to say about waterproofing. The process outlined in the text is entirely reasonable for caulking a normal-size boat, with the quantities proportionally adjusted to accommodate the vast surface area involved. However, there are significant differences from the individual details adduced from the two references above. This part of the tablet is heavily abraded with a number of incomplete lines, but enough remains to clearly see the nature and order of the steps involved, which appear to throw new light on how bitumen was processed for caulking boats.


Here, as before, Atra-hasīs’s 3,600 measures are to be taken seriously. The first step is to work out how much bitumen will be needed to complete the process, and in lines 18 and 19 we are told that Atra-hasīs apportioned one finger thickness of iṭṭû-bitumen, for the inside and outside of the hull. This is where the area calculations we looked at earlier come into their own. As the bitumen will be applied in a uniform layer, one need only work out the area of the boat, multiply it by two for the inside and outside, and then multiply it by the thickness of the coating. However, as the boat itself is one finger thick the work has already been done, and the amount of iṭṭû-bitumen needed is twice the volume of fibre needed to make the hull, which was four šār-and-a-bit, so something over eight šār. This is exactly the sort of calculation a scribe accounting for boat-building materials would have to make, and the sort of problem that would be practised diligently in the scribal school.

Line 20 tells us that the interior cabins have already been coated with one finger thickness of liquid iṭṭû-bitumen, thus focusing our attention on the critical task of waterproofing the hull.


Lines 21 and 22 tell us that indeed eight šār of kupru-bitumen have been loaded into the kilns and one šār of iṭṭû-bitumen is to be poured in as well. That is, we have our two × four šār-and-a-bit, as anticipated above. The eight šār will form the one-finger-thick base coat on the inside and outside of the vessel, while the remaining one šār will be applied as a thin protective top-coat to the outside. Notice, however, that although we are told we need one finger thickness of iṭṭû-bitumen for the inside and outside of the hull, we are actually loading almost entirely kupru-bitumen into the kilns as raw material (as well as a small proportion of iṭṭû-bitumen as a liquid – it is poured in).

This can probably be explained by lines 23, 24 and 25, which read: ‘The iṭṭû-bitumen did not come to the surface [lit. up to me), (so) I added five fingers of lard, I caused the kilns to be loaded … in equal measure.

We interpret this as alluding to the process of fractionation. The kupru-bitumen with fresh bitumen was probably in its native form, solid and containing plant and mineral impurities, and heating it in the presence of oil releases the more-fluid iṭṭû-bitumen, which rises to the surface and can be ‘creamed off’ and used. Much like butter added to a frying pan, the lard transfers the heat to the solid bitumen, preventing it from burning and helping it to melt. The ‘five fingers’ is certainly meant to represent a small quantity, used as a rendering aid, which was then added to all the kilns equally.


We have reached a stage in the bitumen processing where we can assume that the pure liquid iṭṭû-bitumen has been skimmed off the top, leaving only the heavier kupru-bitumen remaining in the kiln. This will have concentrated in it the residue of the plant and mineral impurities from the original raw bitumen. The resulting mastic was presumably used to provide a tough outer layer, similar to that seen in samples of ancient bitumen caulking – which have the appearance of having had tempering agents artificially added. As tamarisk wood is commonly used as firewood, we interpret lines 26 and 27 = ‘I completed … tamarisk wood and stalks’; as referring to increasing the temperature of the fires beneath the kilns in an effort to soften the kupru to make it suitable for application.


Work has now progressed from preparation of the bitumen to its application, and although line 28 is almost totally obliterated, we can tell it refers to coating the interior surface of the hull, by line 29, which can be read as ‘going between her ribs’.


Again, line 30 has been reduced to indecipherable traces, but it must have described covering the exterior surfaces with iṭṭû-bitumen, as this is mentioned in line 31. This base layer is a fine waterproof coat, which must be free from impurities and sufficiently plastic not to crack when the boat flexes. By lines 32 and 33 it is already in place, as a further protective coat is here being applied: ‘I applied the exterior kupru-bitumen from the kilns, using the 120 gur set aside by the workmen.’ This is obviously the remains of the initialkupru-bitumen after all the iṭṭû-bitumen has been extracted. It would form a stiff protective shell over the waterproofing coat of iṭṭû-bitumen.

This order of coatings is the second point which differs from the details from the references in Leemans 1960, which suggest that a crude layer of kupru-bitumen is put on first which is then overlaid with a finer layer of iṭṭû-bitumen to improve it. However, the account suggested here tallies more with the ethnographic accounts of Iraqi reed boat-building given in Ochsenschlager 1992, where the still-hot waterproofing bitumen layer is coated with river mud, which binds to it and forms a strong protective layer. The actual figure on the tablet for the amount of kupru-bitumen used is ‘two gur’, but the nature of Babylonian numbers allows the possibility that this two can be understood as representing any factor of sixty. A coating using two gur would be too thin to be meaningful, and a coat using 7,200 gur would use much more bitumen than we have. Interpreting the two as 120 gur equates to a thickness of exactly one-sixth of a finger when applied to the whole exterior of the Ark. Now 120 gur is equal to one šār, so it must be asked why the quantity reserved by the workmen is not given in this fashion. We believe it to be because – rather than a raw material – it is a finished product gathered from the kiln in vessels more appropriate to measurement in gur.

Another important thing to note is that although – as in the references – the amount of raw kupru-bitumen used (eight šār) was indeed much more than iṭṭû-bitumen (one šār), by the time the bitumen had been cooked and the final products manufactured, these quantities would have been completely reversed, with eight šār of iṭṭû-bitumen being used as opposed to one šār of kupru-bitumen as the dregs. That is, the text suggests that the relative proportions of these types of bitumen is not fixed, but can be altered through a basic industrial process involving heating, much like the relative proportions of ice and water.


The final part of waterproofing and sealing the boat comes in lines 57–8, after a gap in which the Ark is loaded up with animals and supplies. The lines read: ‘I ordered repeatedly a one-finger (layer) of lard for the girmadû out of the thirty gur which the workmen had put to one side.’ As discussed, we consider that the girmadû is the roller-tool for applying the lard, which is the final operation before the boat is, as it were, ready for what lies ahead.

We are grateful to Sir Peter Badge for confirmation that oil is often applied in the construction of traditional guffas, where it can soften and prevent cracking in the outer waterproof layer, the tough coating of kupru in the case of the Ark.

Utnapishti’s Ark

We turn finally to the revealing construction data preserved within Gilgamesh XI. Here the scribes are working with walls at ten nindanu, which are ten times higher than in Atra-hasīs’s Ark. One of the Gilgamesh XI tablets gives the bitumen quantity for waterproofing at nine šār, transmitting correctly the original Old Babylonian quantity and not adjusting it in terms of the ‘new’ walls. (The other gives six.) However, this nine šār of bitumen is to serve for the whole cubic Ark. This means that if Utnapishti’s craft is waterproofed with a standard thickness of one finger for the bitumen, simple calculation shows that there would not be enough to do the interior at all, and the exterior could only be waterproofed to a height of 6.5 nindan up the walls, incredibly close to the 6.66 or two-thirds that the ‘oiling’ by the girmadû covers.

To us this means that the Gilgamesh editor has used the given height of the walls and the given quantity of bitumen to calculate the coverage this would provide, and then edited this new data into the story. Otherwise the appearance of the ‘two-thirds’ here is rather hard to explain. Unfortunately, in Gilgamesh XI the thirty gur of lard for the girmadûs has ended up as two šār – a completely unfeasible amount – and here the scribe has been unable to make sense of this.

Ut-Napishtim’s Ark coated with bitumen to about 2/3 its height.

If you find an error please notify us in the comments. Thank you!