Decoding Medieval Texts with AI and Multispectral Analysis

Medieval manuscript scholars are increasingly turning to advanced technology to uncover the hidden histories of ancient texts. By combining multispectral ink-composition analysis with artificial intelligence designed for hand-scribe clustering, researchers can now map the exact workflows of monastic scriptoria. This article explores how these non-invasive methods identify the chemical makeup of inks and pigments, differentiate between individual scribes, and reveal how medieval monks collaborated to copy and illuminate priceless manuscripts.

Multispectral Ink-Composition Analysis

Medieval scribes primarily used iron gall ink, a mixture of oak galls, iron sulfate, and gum arabic, alongside vibrant mineral and plant-based pigments for illuminations. Because every monastery—and sometimes each individual batch of ink—had a slightly different recipe based on local materials, the chemical composition of the ink acts as a historical fingerprint.

Multispectral imaging (MSI) captures photographs of manuscript pages under varying wavelengths of light, including ultraviolet, visible, and near-infrared spectra. Different chemical elements absorb and reflect these wavelengths in distinct ways. Scholars pair MSI with X-ray fluorescence (XRF) and Raman spectroscopy to read these reflections without taking physical samples that would damage the parchment.

By analyzing the chemical ratios of trace metals like copper, zinc, and iron in the ink, scholars can determine:

Geographic Origin: Specific mineral impurities link inks to local water sources or regional trade routes.
Chronology: Changes in ink batches across a single manuscript show where work paused and resumed.
Corrections and Additions: MSI easily separates original text from edits made decades later by different hands using different ink formulas.

Hand-Scribe Clustering Artificial Intelligence

Traditionally, paleographers identified individual scribes by manually comparing letterforms, a slow process prone to human subjectivity. Today, hand-scribe clustering uses machine learning and computer vision to automate and refine this process.

Researchers train convolutional neural networks (CNNs) on high-resolution scans of manuscripts. The AI breaks down the handwriting into measurable data points, analyzing micro-features that the human eye might miss. These features include:

The exact angle and curvature of pen strokes.
The spacing between specific letters and words.
The proportions of ascenders and descenders (the parts of letters that extend above or below the main body).
The pressure applied to the parchment via the quill.

Using unsupervised learning algorithms, the AI groups similar writing styles into clusters. Each cluster represents a distinct scribe. This allows scholars to feed an entire digitized manuscript into the system and quickly see how many different monks worked on the text, even if those monks were trained to write in the exact same uniform scriptorium style.

Reconstructing the Monastic Scriptorium

When scholars combine chemical ink data with AI handwriting clusters, they get a clear, page-by-page reconstruction of how a monastic scriptorium operated.

First, the division of labor becomes visible. A single manuscript was rarely the work of one monk. AI clustering frequently reveals that manuscripts were divided into "quires" (booklets of folded parchment) and distributed among several scribes to copy simultaneously. For example, Scribe A might copy pages 1 to 16, while Scribe B copied pages 17 to 32.

Second, the hierarchy of the scriptorium emerges. By matching AI handwriting clusters with ink shifts, scholars often observe a master scribe writing the first few lines of a page to set the style, followed by an apprentice taking over to complete the rest.

Finally, this dual approach clarifies the illumination process. Multispectral analysis of the pigments—such as lapis lazuli, vermilion, and gold leaf—combined with ink analysis shows the sequence of production. Scholars can see how scribes left blank spaces for the rubricators (who added red lettering for headings) and the illuminators (who painted the illustrations). In many cases, the data proves that a manuscript was copied in one monastery using local iron gall ink, and then shipped to an entirely different, wealthier monastery with access to exotic pigments for the final illumination.