Two scribes so similar only a computer could tell them apart
Nos confins do deserto de Judá, onde os manuscritos mais antigos da Bíblia hebraica dormiam há dois milénios, a inteligência artificial veio confirmar o que o olho humano apenas suspeitava: duas mãos distintas, de formação tão semelhante que só um algoritmo as conseguiu separar, escreveram o Grande Rolo de Isaías. A equipa da Universidade de Groningen, liderada por Mladen Popovic, publicou em abril de 2021 uma análise computacional que transforma a paleografia numa ciência quantitativa, devolvendo rostos anónimos à produção dos Manuscritos do Mar Morto.
- Durante décadas, a questão de quantas mãos escreveram o Grande Rolo de Isaías permaneceu sem resposta definitiva, presa entre a intuição dos especialistas e a ausência de provas mensuráveis.
- A escrita dos dois escribas é tão semelhante que o olho humano, mesmo o mais treinado, não conseguia distingui-los com certeza — criando uma tensão entre o que a tradição académica suspeitava e o que podia ser demonstrado.
- Investigadores holandeses desenvolveram redes neurais artificiais capazes de preservar e analisar os traços de tinta originais, medindo tanto a textura dos traços individuais como a forma completa das letras.
- A análise das cinquenta e quatro colunas revelou dois grupos distintos e não aleatórios, com uma transição clara a meio do documento, confirmada por três métodos independentes.
- O segundo escriba mostrou maior variação estilística do que o primeiro, mas a semelhança entre ambos sugere uma origem ou formação comum — um detalhe que humaniza a produção deste manuscrito milenar.
- A metodologia abre caminho para identificar escribas individuais em outros textos de Qumran, convertendo a paleografia numa disciplina reprodutível e estatisticamente rigorosa.
Há muito que os estudiosos suspeitavam que o Grande Rolo de Isaías, um dos manuscritos bíblicos mais antigos alguma vez descobertos, não era obra de uma só mão. A escrita parecia quase uniforme a olho nu, mas algo na sua textura sugeria uma segunda presença. Em abril de 2021, uma equipa da Universidade de Groningen, liderada por Mladen Popovic, publicou na revista PLOS One a confirmação que séculos de estudo apenas podiam intuir: dois escribas distintos escreveram este texto, com mãos tão semelhantes que só um computador os conseguiu separar.
O Rolo de Isaías é um dos Manuscritos do Mar Morto descobertos há mais de setenta anos em grutas próximas do Mar Morto, em Qumran. Com mais de dois mil anos, contém uma das cópias mais antigas da Bíblia hebraica e estende-se por cinquenta e quatro colunas de texto escritas em couro ou papiro. Para analisá-lo, a equipa incluiu Lambert Schomaker, especialista em inteligência artificial e reconhecimento de escrita manual, que desenvolveu algoritmos para separar a tinta do fundo e redes neurais capazes de preservar os traços originais de tinta — registos diretos do movimento muscular de cada escriba.
A análise combinou características texturais, como a curvatura dos traços individuais, com características alográficas, que examinam a forma completa das letras. O resultado foi inequívoco: as cinquenta e quatro colunas dividem-se em dois grupos distintos, não distribuídos aleatoriamente, com uma transição a meio do documento. Um exame detalhado da letra aleph nas primeiras e últimas vinte e sete colunas confirmou diferenças mensuráveis entre os dois escribas.
Embora estudiosos anteriores já tivessem sugerido, com base em aspetos físicos do rolo, que um novo escriba assumia a partir da coluna vinte e sete, a teoria permanecia contestada por falta de evidências quantitativas. Popovic e a sua equipa mudaram isso. Descobriram ainda que o segundo escriba mostrava maior variação estilística do que o primeiro, embora ambos mantivessem mãos notavelmente semelhantes — sugerindo uma formação ou origem comum. A metodologia promete agora abrir novos caminhos para identificar escribas individuais noutros textos de Qumran, transformando a leitura da escrita antiga numa ciência reprodutível e estatisticamente rigorosa.
Scholars have long suspected that the Great Isaiah Scroll, one of the oldest biblical manuscripts ever discovered, was the work of more than one hand. The document's writing appeared almost uniform to the naked eye, yet something about its texture suggested a second presence. Now, researchers using artificial intelligence have confirmed what centuries of study could only hint at: two different scribes wrote this ancient text, their hands so similar that only a computer could tell them apart.
The finding comes from a team at the University of Groningen in the Netherlands, led by Mladen Popovic, who published their work in the journal PLOS One in April 2021. They focused their analysis on the Great Isaiah Scroll, one of the Dead Sea Scrolls discovered more than seventy years ago in caves near the Dead Sea, primarily at Qumran. These scrolls, now over two thousand years old, contain the oldest known copies of the Hebrew Bible and numerous ancient Jewish texts. The Isaiah manuscript consists of fifty-four columns of text written on leather or papyrus—a substantial document that would have required considerable time and labor to produce.
The research team included Lambert Schomaker, a computer science and artificial intelligence professor who has spent years developing techniques to teach computers to read handwriting and understand how physical factors like pen grip influence writing patterns. The approach was methodical. First, the researchers created an algorithm to separate the ink from its background, then developed an artificial neural network capable of preserving the original ink traces made by the scribe more than two millennia ago. This precision mattered enormously, Schomaker emphasized, because ink strokes are direct recordings of muscle movement and are unique to each person.
The analysis examined two types of features: textural characteristics, which measure things like the curvature of individual strokes, and allographic features, which analyze complete letter forms. When the team applied these measurements to all fifty-four columns, a clear pattern emerged. The text divided into two distinct groups that were not randomly distributed across the scroll but clustered together, with a transition occurring roughly halfway through the document. A second analysis using letter-fragment patterns confirmed the presence of two different writers. The third phase involved a detailed visual examination of the letter aleph in the first and last twenty-seven columns, revealing measurable differences between them.
Previous scholars had suggested, based on physical aspects of the scroll and text positioning, that a new scribe took over after column twenty-seven. But this theory remained contested, lacking the kind of quantitative evidence that could silence debate. Popovic's team changed that. The second scribe, they found, showed more variation in his writing than the first, though both maintained remarkably similar hands—a detail suggesting they shared common training or origin. "Instead of basing judgment on impressionistic evidence, we can now show, with the computer's help, that the separation is statistically significant," Popovic explained.
The implications extend far beyond this single manuscript. The team believes this computational methodology opens a new pathway for analyzing other Qumran texts, potentially allowing researchers to identify individual scribes and observe how each one worked across multiple documents. It represents a shift from subjective assessment to measurable, reproducible analysis—a way of reading ancient handwriting that the naked eye, no matter how trained, could never achieve alone.
Notable Quotes
Instead of basing judgment on impressionistic evidence, we can now show, with the computer's help, that the separation is statistically significant.— Mladen Popovic, University of Groningen
The Hearth Conversation Another angle on the story
Why does it matter that two people wrote this scroll instead of one?
Because it tells us something about how these texts were produced. If one person wrote it, we learn about their discipline and stamina. If two people did, we're seeing a workshop, a division of labor, maybe even a training relationship.
The article says their handwriting was so similar that only a computer could tell them apart. Doesn't that seem almost too close to be coincidence?
That's exactly what the researchers thought. Two scribes with nearly identical hands suggests they learned from the same teacher, or worked in the same scriptorium for years. They weren't strangers picking up where someone left off.
How does a computer actually detect the difference if human eyes can't?
It measures things we can't see—the exact curve of a stroke, the pressure patterns in the ink, the tiny variations in how letters connect. A human sees "this looks like the same hand." A computer counts thousands of micro-differences and finds the statistical signal underneath.
The second scribe had more variation in his writing. What does that tell you?
Maybe he was younger, less practiced. Or maybe he was more tired by the end of the document. Or simply less disciplined. But the fact that we can measure it now—that's new. We're not guessing anymore.
What happens next with this technique?
They want to apply it to other scrolls from Qumran. Imagine being able to track individual scribes across multiple manuscripts, seeing which texts they worked on, how their hands changed over time. You could map an entire scriptorium's production.