In a merchant town in central Anatolia, in the first half of the 19th century BCE, a woman named Anna-anna addressed a letter to her husband about money. It was never opened. The tablet, catalogued as Kt 94/k 1150, is still inside the clay envelope she rolled her seal across eight times, and until recently reading an Old Assyrian letter meant breaking that envelope and ruining the seal impressions. A portable X-ray scanner carried into the museum has now lifted the tablet’s surface out of the clay, and an Assyriologist has read it.
The letter is a debt complaint. Her husband Ennum-Aššur, a merchant, was away trading and had sent instructions to collect silver a man named Laqēpum owed him. Every time she went, Laqēpum said he had already handed it to a third man, Kudātum. She told him she would produce witnesses to testify against him if he had not. He would settle up with Ennum-Aššur in person, he said, when Ennum-Aššur got home.
Then she leaves the debt and summons her husband: “If within 10 days you have not arrived, as for your departure, stand up and come.” And then the line that needs no translator’s footnote. “I hear bad words all the time!”
The reading was published in May 2026 in npj Heritage Science by a team led by the Assyriologist Cécile Michel and the physicist Christian Schroer. The machine was assembled on site in the Museum of Anatolian Civilisations in Ankara.
Old Assyrian merchants wrapped their letters in a second skin of clay. The outside carried almost nothing: the sender’s name, the addressee’s name, and the sender’s cylinder seal rolled across the surface. Anna-anna’s carries three lines of text and her seal twice on the obverse, twice on the reverse and once on each edge.
A letter envelope did three jobs at once. It protected the tablet on a journey of roughly a thousand kilometres from Assur, in what is now northern Iraq, to the merchant colony at Kanesh. It authenticated the sender. And it kept the contents private, which is the whole point of the design.
So an intact letter envelope is a failure. The paper’s own reading is that a letter still sealed is a letter that never arrived, either because it was not sent or because it was not delivered. A sealed contract envelope implies something else: on the paper’s account a contract envelope was broken once the contract was completed, so one still intact is a contract that was still running. Several hundred envelopes of both kinds were found still intact in the merchants’ houses at Kültepe, the modern name for Kanesh.
The cost of looking inside
A small number of sealed envelopes were cracked open in the last century so that Assyriologists could get at the tablet, and the miniature scenes rolled into the outside were irreparably damaged, putting them beyond study. For several decades now the envelopes have no longer been opened. That policy preserves the seals and freezes the letters.
Computed tomography is the obvious escape, and not new to cuneiform: Old Assyrian contracts from a Leiden collection were scanned at TU Delft, and the British Museum has a micro-CT system of its own. The obstacle is logistics. One of the two commercial machines the paper names has a mass of almost two tonnes; the other reaches up to six and a half. Museums are only rarely equipped with one and do not usually let artefacts leave the building.
The Ankara collection makes that concrete: tablets there can only be studied at a few workbenches in the corridor outside the archive, reached by a narrow staircase. If the tablets cannot go to the scanner, the scanner has to fit up the stairs.
A scanner that comes apart into eight pieces
ENCI, for Extracting Non-destructively Cuneiform Inscriptions, was designed and built over about four years at the DESY research centre in Hamburg, and splits into eight modules. Most weigh around 50 kilograms and can be carried by two people; the server rack holding the control and computing units is about 100 kilograms and is typically handled by four. The whole assembly weighs roughly 420 kilograms, and goes from parts to working order on site in about two hours.
The radiation requirement is what drives the weight. Penetrating clay objects of the size this scanner was built for needs X-rays well above 100 keV, so the tube runs at up to 180 kV, and the scanner has to be safe standing in a public corridor. The team set that as a dose rate in the scanner’s vicinity never exceeding 0.5 microsieverts per hour, a ceiling chosen, they write, to stay below the 1 millisievert effective dose limit that the general public is typically exposed to over a year from all other external sources. It is CE certified, approved under the German Radiation Protection Act, and holds full radiation-protection approval in Türkiye for mobile operation for the next five years.
One unglamorous trick does a lot of the work. Clay swallows the low-energy end of the spectrum, which distorts the reconstruction badly: without filtering, the reconstructed attenuation at the centre of a replica tablet comes out almost a factor of two lower than at its rim. Putting two millimetres of copper in front of the source throws away the part of the beam that was barely going to get through anyway, and the variation drops to about ten per cent.
The scan is the short half
A scan is typically 1,440 images taken over a full rotation, finishing in under thirteen minutes, with the reconstruction of a full volume taking under fifteen minutes on the machine’s own server. At full size the volume runs to about 73 gigabytes. The paper deposits photographs, volume data and 3D geometry files publicly for five of the tablets; in Anna-anna’s own deposited record the envelope and the tablet are separate geometry files.
Then the harder half begins, because a tomogram is a block of densities, not a page. The software has to find the boundary between two objects usually made of the same material, separated by an air gap the paper calls extremely thin and sometimes absent altogether: as the envelope’s clay dries it follows the tablet’s contours so closely that the wedges pressed into the tablet show up in relief inside the envelope. Only once tablet and envelope are pulled apart in software does anything become readable.
The number to hold onto is the resolution. The instrument was specified to resolve better than 50 micrometres, and in principle, depending on the imaging geometry and magnification, it can reach 15 to 40. The reconstructions in this paper came in at about 65 micrometres, plus or minus 10, and the authors say plainly why: the scans were run fast, and photon noise sets the limit. Speed was the point, since archive opening hours are short. The paper does not treat that as a shortfall and never sets the achieved figure against its own objective. The texts were read either way, and the published reconstructions sit coarser than the specification.
What else the scans turned up
Anna-anna’s letter is one of five tablets the paper presents from the Ankara trip, and one of three whose hidden text it transliterates and translates. In another envelope, a bulge on the outside turned out to be a second, smaller tablet tucked against the first: an Old Assyrian writer who ran out of room could shape an extra oval tablet, a ṣibat ṭuppim, and add up to a dozen more lines. Loose ones have turned up before, but once an envelope was opened in antiquity there is generally nothing to say which letter they belonged to. This one is still in place against its parent letter, which announces a consignment of textiles carried on two black donkeys; the paper adds that the text on the additional tablet is not clearly linked to the text on the main one. The practice, the paper says, is attested only in the Assyrian archives at Kanesh.
A third envelope carried a letter of instruction about a contract, with a phrase that repays the scan. The tablet inside tells its addressees to hand a debt document back to the debtor, who is one of them, so that he can cancel it, or, in the scribe’s own word, kill it. In the same object the scan found an inclusion trapped in a layer of the envelope, which the authors say appears to be a seed. Andrew Fairbairn, the archaeobotanist at the Kültepe excavation, identified it: “This is a barley seed, clearly visible in the overall shape with the chisel-like ends, dorsiventrally flattened shape and shallow ventral furrow.” He adds that it is probably hulled barley.
Two more findings came from the clay rather than the writing. In one tablet from the house of Elamma, excavated in 1991, the clay of the tablet and the clay of the envelope are completely different, which the authors suggest means the envelope went on much later, probably elsewhere. And the envelopes turn out to be built from several folded layers rather than a single sheet, which is not what an earlier reconstruction based on envelope fragments had proposed.
The edges of the method
How many tablets the Ankara trip covers is slippery. The abstract says around fifty encased tablets and the introduction around fifty encased letters and contracts; the mission account says 64 tomograms of 48 tablets and other objects over three weeks. The counts are close, but “tablets and other objects” is a wider category than “encased tablets”, so the number of sealed envelopes scanned may be lower than either suggests.
The multilayer result shows what this technique does and where it stops. It contradicts a previous account of how envelopes were made, and the authors do not declare the question settled: they write that further experimentation is required to confirm the observations. That is a hypothesis upgraded by imaging, not a manufacturing process reconstructed.
Five tablets is a demonstration, not a survey, against several hundred intact envelopes from Kültepe alone and the half a million to a million cuneiform artefacts from the region. The project has kept going: its page at Hamburg’s Centre for the Study of Manuscript Cultures reports two long Ankara missions, in 2024 and 2025, and over one hundred and fifty sealed envelopes scanned between them, along with a stop at the Kayseri Archaeological Museum and a collection in Jena. Only a fraction is published.
The scanner also does not read anything by itself. It produces surfaces. The signs on them still have to be transliterated and translated by someone who knows Old Assyrian, and the paper is candid that the extra tablet in the second envelope was hard to separate in software, because the main tablet’s signs are pressed into its back in negative. Where the imaging fails it will presumably fail quietly, as a surface that never resolves into a sign.
The narrow thing the machine was built to do is not in doubt. A letter that in this museum could previously be read only by destroying the seals that are the reason not to open it can now be read with them intact, in the building where it lives.
There is a more general bargain underneath all this. Whatever people do to keep a document from being read is usually the same thing that carries it intact into a future they were not imagining, and the two cannot be separated at the time of writing. The price of collecting on the second used to be the first. On the evidence so far, that price is starting to look optional.