How a Sealed Brienne Letter Was Read Without Opening It

Packet DB-1627 was dated 1697-07-31, and it remained physically sealed when researchers produced readable images of the writing inside it. That pairing sounds impossible only if reading is assumed to require breaking a seal and spreading paper flat. Here, the documented achievement is more precise: an X-ray scan and a computational process created a view of writing without opening the packet. But a readable image is not an algorithmic understanding of language, nor a finished account of the people named in the letter.

The Brienne Collection makes that dividing line important from the outset. Its catalog contains 3,148 items, including 577 unopened letterpackets, within an early-modern postmaster’s trunk that also holds delivered, undelivered, and opened letters. Those figures describe the collection, not a count of packets fully read by the study. Yet they show why the question extends beyond one message: each sealed object may preserve writing alongside folds, tucks, slits, and seals whose arrangement cannot be reduced to a transcription alone.

DB-1627 gives the method a human-sized test rather than an abstract technical demonstration. Jacques Sennacques sent it from Lille to his cousin Pierre Le Pers, a French merchant in The Hague, asking for a certified copy of the death notice of Daniel Le Pers. The request is specific enough to feel close at hand, but the sealed packet kept the letter’s physical arrangement intact. How can researchers examine words on internal paper surfaces while leaving the object that carried them in that condition?

The sealed packet DB-1627 begins with a question of access

MIT News describes the collaboration’s non-destructive scanning; MIT CSAIL details its 3D outputs and keeps historical interpretation outside the algorithm. Letterlocking research approaches a sealed letterpacket as more than a container around a message. Folds, tucks, slits, and seals can preserve evidence of how a letter was secured and handled. That makes DB-1627 unusual in a productive way: the same arrangement that stopped a casual reader from reaching the writing also became part of what scholars did not want to destroy. Opening it would solve one access problem while ending the chance to study the closures in their original relationship.

The Brienne Collection’s still-sealed packets therefore pose a double demand. Researchers want the language inside, but they also want the packet’s folded configuration available for later study. A conventional opening would alter the relationship among paper layers and closures that letterlocking examines. Virtual unfolding offers a different route: it can produce a digital surface for reading while the original packet stays sealed, retaining its physical configuration for later questions about how that message travelled as an object.

DB-1627 is the documented worked example because its message could be read while the packet remained physically closed. The 2021 open-access Nature Communications paper identifies the sender, recipient, route between Lille and The Hague, date, and request for a certified death notice. Those details come from reading the rendered letter; they do not mean software independently translated the language or interpreted the family circumstances. Human paleographical reading and historical judgment begin after the computational output has made the writing available for examination.

That boundary changes the drama of the letter. The discovery is not a machine uncovering a family story by itself; it is a method making a sealed writing surface accessible in a form people can inspect. The paper validates its pipeline against an already opened packet before applying it to sealed Brienne packets, providing a comparison point rather than a theatrical claim of effortless access. The larger collection also prevents an easy extrapolation: hundreds of unopened packets exist, but the study does not say every one has yielded a readable letter.

The paper describes difficult folds and touching paper layers as continuing challenges. A scan-derived rendering remains a transformed representation, and its legibility can depend on whether computational stages can separate and reconstruct the surfaces involved. This is the odd condition of DB-1627: its seal was not simply an obstacle waiting to be removed, but a physical state the method had to respect while finding another route to the writing. The object’s survival and the message’s visibility had to be pursued together.

To do that, the process cannot begin with a flat photograph of the outside. The paper describes X-ray microtomography, volumetric scan processing, segmentation of paper layers, geometric reconstruction, and flattening or rendering as separate stages. Each stage addresses a different difficulty, from locating paper within the scanned packet to presenting a surface on which marks can be read. The next question follows directly from that sequence: what can an X-ray volume retain about layered paper that an exterior photograph cannot show?

X-ray microtomography turns folded paper into a volume

When a sealed Brienne letterpacket enters an X-ray microtomography scan, the team is not acquiring a photograph of its message. They are acquiring a three-dimensional volume in which the packet’s paper and folded form can be computationally examined. The 2021 open-access paper presents that scan as the beginning of a pipeline, not the instant at which a letter becomes legible. That distinction matters: physical acquisition leaves the closed object intact, while every later view is derived from measurements taken through it. The first difficulty is spatial—locating paper that has been bent, tucked, and sealed into a compact packet.

X-ray microtomography supplies the volume, but it does not announce which portion belongs to which sheet. The paper describes a first step in which the scanned packet is separated from its surroundings before its folds are reconstructed. At this point, the scan is a measured object rather than a flat document; useful information lies through its depth as well as across its visible outline. We should resist picturing a machine that simply sees through paper and delivers a page. It provides data from which individual layers must still be identified, especially where folded material lies close together.

Segmentation is the first decisive sorting stage. In the authors’ workflow, paper layers are separated within the volumetric data so the team can follow their surfaces rather than confuse one folded region with another. The procedure does not loosen a tuck or break a seal; it assigns a digital boundary to material already captured in the scan. This preserves a crucial part of letterlocking evidence. The physical arrangement—the folds, tucks, slits, and seals by which a letterpacket was secured—stays in place, even while a computational model begins to distinguish the surfaces enclosed within it.

That separation is not a magic X-ray reading. The methods description lists volumetric scan processing and paper-layer segmentation as distinct stages, and that division clarifies what the result cannot mean. A segmented layer is a proposed digital reconstruction of a surface inside a closed object, not a newly opened sheet in a researcher’s hand. Where layers touch or folds are especially difficult, the paper acknowledges challenges that can limit the result. Published supporting data and code references allow the workflow to be examined, but they do not make difficult geometry disappear.

Once surfaces have been identified, geometric reconstruction takes over. The pipeline reconstructs the folded form so a computational surface retains a relationship to the packet’s bends rather than being treated as a loose rectangle from the start. This step is necessary because writing surfaces can face in different directions inside a sealed object. A reader who jumps straight from scan to decoded letter misses the intervening work: the model must carry each recovered surface through its folds before flattening can make a coherent view. Only then can the mapped faces be laid out without suggesting that the physical packet itself has changed.

The reconstructed geometry also marks a limit on the claim. Its output is a transformed representation derived from the scan, not a substitute for the original letterpacket or a guarantee that every mark can be recovered. It lets a human reader inspect a digital rendering while the closed original retains its physical configuration for future study. For the Brienne Collection, that restraint has special force: its postmaster’s trunk contains delivered, undelivered, opened, and still-sealed letters, and sealed packets preserve information in their folds as well as their words.

Mapping layers explains how separate paper surfaces can be located; it does not yet place their writing into a readable order. The documented pipeline then reconstructs and renders selected surfaces for examination. Those outputs remain digital representations rather than a physical opening, automatic translation, or settled historical interpretation. DB-1627 gives that distinction a dated human scale.

The Brienne Collection's folded layers are flattened on screen

To flatten a sealed letter is not to press it flat, loosen a seal, or alter a fold. The 2021 Nature Communications paper describes a computational route from X-ray microtomography images to a digital representation of the packet’s writing surfaces. The physical object remains sealed while its scanned form is handled as volume data. That distinction matters: the visible result is made from measurements of the folded letter, not from a newly opened sheet placed on a desk.

The process begins with physical acquisition: X-ray microtomography produces the scan from which the packet can be examined computationally. The Nature Communications paper then separates the scanned packet from its surroundings and processes its volume. This is not yet a reading of a letter. It is the creation of a three-dimensional data object whose internal paper layers must be distinguished before any words can be brought into view. For related archive context, compare How Virtual Unwrapping Reads the and How Multispectral Imaging Recovered the.

Those layers are the difficulty. A sealed letterpacket is not simply a page rolled into a convenient shape; its folds, tucks, slits, and seals helped secure it. Letterlocking research on the Brienne Collection treats those features as evidence of handling and protection rather than disposable wrapping. The strange part is that the same folds that conceal the writing also carry information about how the object was made secure. A method that exposes text by destroying those features would lose part of what made the letter worth studying.

After segmentation, the pipeline reconstructs the folded geometry of the paper layers. The Nature Communications methods and data availability record describes geometric reconstruction as distinct from both scan processing and later flattening or rendering. In practical terms, the system must estimate a usable surface through a form that has been bent back on itself. If layers touch closely or folds are difficult, that task can remain challenging. A screen image may look continuous while depending on a complicated reconstruction underneath it.

Flattening follows the reconstruction. The computational process maps a selected writing surface into a form that can be viewed more like a page, then renders readable images from that transformed surface. The result can reveal writing without changing the sealed original, but it is not the original object transported intact onto a monitor. It is an image derived from the scan and the sequence of transformations applied to it. Each stage has done a different job: acquisition captured the packet, segmentation separated layers, reconstruction described their shape, and flattening arranged a surface for viewing.

This division keeps the word “read” from doing too much work. A rendered surface can make letterforms available, yet historical transcription still requires a person to identify what those letterforms say, and interpretation asks further questions about their meaning and context. Supporting data and code references published with the method allow the workflow itself to be examined, but they do not turn every difficult fold into certainty. Once the surface becomes visible, the next question is not whether the machine has understood the letter, but what a researcher can responsibly make of the writing it has rendered.

Jacques Sennacques's letter remains closed after it is read

Reading a rendered image of a folded page is not the same as opening DB-1627, and that distinction governs every claim about it. The sealed packet is dated 1697-07-31; the Nature Communications case study identifies Jacques Sennacques in Lille as sender and Pierre Le Pers, a French merchant in The Hague, as recipient. A readable rendering was recovered while the packet stayed physically sealed. We can follow a dated communication across those named places, but the object has not become an ordinary opened letter whose folds no longer require attention.

In the case study, Sennacques asks his cousin for a certified copy of the death notice of Daniel Le Pers. The request is specific enough to prevent the letter from becoming a vague technical display: it asks for a certified copy of Daniel Le Pers's death notice across the distance between Lille and The Hague. Yet the computational process did not supply a theory of the relationship, the circumstances of Daniel Le Pers's death, or the sender's private feelings. Those questions begin only when human readers confront the rendered words and decide how much their handwriting can securely yield.

The rendered surface gives historians a new view to read while the sealed packet remains physically intact. It does not translate the letter or decide its historical meaning.

MIT CSAIL reports that the 3D analysis separates thin paper layers and produces folded and flat reconstructions, writing-surface images, and crease patterns. Those outputs expand access without claiming that every mark or every packet has been recovered, and historical reading remains a separate human task.

DB-1627 also acquires context from the Brienne Collection, an early-modern postmaster's trunk containing delivered, undelivered, opened, and sealed letters. Its catalog contains 3,148 items, including 577 unopened letterpackets. Those figures describe the collection, not the number of packets fully read through this method. The difference is essential: one documented success offers a way to approach further packets, but it cannot become a claim that every sealed Brienne letter has yielded its contents. The scale of the trunk makes restraint more important, not less.

The methods and data availability material identifies supporting data and code references for examination alongside the computational workflow. Letterlocking research supplies a second kind of attention: folds, tucks, slits, and seals can preserve evidence of how correspondence was secured and handled. Keeping DB-1627 closed retains that physical configuration for later examination while a digital rendering permits access to writing on its surfaces. The rendered letter and the folded packet therefore answer different questions. Neither cancels the other, and neither alone supplies the whole history of the message.

Jacques Sennacques's letter can now be read as a request sent from Lille to The Hague on 1697-07-31, yet its folded body has not been sacrificed to make that reading possible. On a screen, words from the packet become available; in the Brienne Collection, the same packet remains shut, with its sealed physical configuration intact for later examination. A letter that once had to be broken open to speak can now be heard through its folds while those folds remain intact.

Frequently Asked Questions