The Punch Card Prophet: The Unlikely Origins of Computational Humanities

The cold stone of the Sapienza University of Rome’s basement archives seeped through Rosario Romeo Monteleone’s wool coat on the November day in 1949 when he first pitched his impossible idea to the local IBM representative. For 18 months, the 32-year-old Italian literature scholar had been manually tallying every instance of the word *amore*—love—across Dante Alighieri’s 14,233-line *Divine Comedy*, cross-referencing each occurrence with its canticle, stanza, and surrounding context to build a concordance: a systematic alphabetical index of every word in the text paired with its usage examples, a tool that had been the holy grail of Dante scholarship for nearly 700 years. His fingers were calloused from turning parchment, his eyes strained from poring over microfilm, and he estimated he was only 30% done. At his current pace, the concordance would not be published in his lifetime.

The IBM representative stared at him like he was mad. The machine he was offering was the IBM 405, a 2.5-ton punch card tabulator invented in the 1930s for U.S. census data processing, designed to sort and count perforated cardboard cards at a rate of 100 cards per minute. It had never been used for anything resembling literary analysis. Monteleone begged. He promised IBM would get full credit in the concordance’s preface, that the project would prove the machine’s utility far beyond census bureaus and payroll departments. After three weeks of haggling, the representative agreed to lend him the machine for three months, on the condition that Monteleone’s team of three graduate students would learn to operate it themselves.

The first week was a disaster. The students punched 12,000 cards with Dante’s text, one word per card, only to have the machine jam 47 times in the first hour of operation. But by the second week, they had worked out the kinks. When the first printout of *amore*’s frequency arrived, Monteleone’s hands shook: the machine counted 1,249 instances, while his manual tally had come to 1,247. He spent two days rechecking the original folios and found he had missed two obscure uses in the *Paradiso*’s final canto. The machine was not just faster—it was more accurate. By March 1950, Monteleone’s team had processed all 14,233 lines of the *Divine Comedy*, building the first full concordance of a major literary work created with a punch card machine. When the 547-page volume was published in 1951, it sold out in three months, and Dante scholars across Europe were suddenly able to trace the semantic evolution of words like *virtù* (virtue) or *peccato* (sin) across the entire poem in days, rather than years. Monteleone had not just built a reference tool—he had opened a door no one knew existed.

The Prehistory of Computational Humanities: Busa, Tabulators, and the First Concordances

Mentoneleone’s project was not, in fact, the first foray of punch card technology into the humanities. Three years earlier, in 1946, an Italian Jesuit priest named Roberto Busa had begun a far more ambitious project: building a concordance of all 13 million words in the works of Thomas Aquinas, the 13th-century theologian whose writings formed the core of Catholic intellectual tradition. Busa had tried to build the concordance manually, but after two years of work, he realized he would need 30 more years to finish it. He wrote to Thomas J. Watson, the president of IBM, asking for access to the company’s latest tabulating machines. Watson, intrigued by the audacity of the request, agreed to sponsor the project, providing Busa with free access to IBM’s New York City lab and later its machines in Milan.

Busa’s project would take 34 years to complete, culminating in the 101-volume *Index Thomisticus*, published between 1969 and 1980. It required more than 10 million punch cards, which when stacked reached a height of 3 meters, and processed 1.2 million pages of Aquinas’s Latin texts. Unlike Monteleone’s project, which focused on a single literary work, Busa’s concordance was designed to let scholars trace the use of theological concepts across Aquinas’s entire corpus: they could see, for example, every instance of *gratia* (grace) in Aquinas’s commentaries on the Bible, his *Summa Theologica*, and his sermons, side by side, to map how his understanding of the concept evolved over his career. The *Index Thomisticus* was not just a reference tool: it was the first proof that computational methods could answer questions that were simply unanswerable with traditional close reading, no matter how many scholars were assigned to the task.

These two early projects—Mentoneleone’s *Divine Comedy* concordance and Busa’s *Index Thomisticus*—set the template for the first two decades of what would later be called digital humanities. Both were driven by a single, practical problem: the impossibility of manually processing large textual corpora to answer specific research questions. Both used the only computational technology available at the time: punch card tabulators, later replaced by mainframe computers in the 1950s and 1960s. And both were rooted in a humanistic problem, not a technical one: Monteleone wanted to understand Dante’s use of love as a thematic device; Busa wanted to map Aquinas’s theological development. They were not computer scientists experimenting with humanities data: they were humanists who adopted computers as a tool to solve problems their training had not prepared them to solve alone.

The field that grew out of these early experiments was initially called *humanities computing*, a term that emphasized its identity as a subfield of computing, not of the humanities. The first academic conference dedicated to the field was held in 1964 at Harvard University, organized by Anthony Kenney, a classicist who had used IBM mainframes to build a concordance of Greek tragedy. Kenney also founded Harvard’s first humanities computing lab, which used the university’s IBM 7090 mainframe to build concordances of Shakespeare’s complete works and John Milton’s *Paradise Lost*. The lab’s output was used by hundreds of literary scholars across the U.S. and Europe, and it established a standard for digital textual analysis that would persist for decades.

In the 1970s, the field expanded beyond textual concordances to include other forms of computational analysis. In London, Susan Hockey, a classicist and computer scientist, founded the “Computer-Assisted Literary Studies” project at King’s College London, which built the first large-scale concordance of Old English texts, including *Beowulf* and the entire corpus of Anglo-Saxon poetry. Hockey also founded the Oxford Text Archive in 1976, a repository of digitized literary and linguistic corpora that remains one of the most important resources for digital humanists today. In the U.S., the first journal dedicated to the field, *Computers and the Humanities*, was launched in 1966, and by the end of the 1970s, there were more than 20 universities offering courses in humanities computing, most of them focused on teaching scholars how to use mainframe computers to build concordances and run basic statistical analyses of texts.

For the first two decades of its existence, humanities computing was a niche, technical field, dominated by scholars who were as comfortable writing FORTRAN code as they were close reading medieval poetry. Most humanists viewed it with suspicion: in a 1978 survey of members of the Modern Language Association, 72% said they believed “computers could never replace the intuitive judgment of a trained literary scholar.” The field’s focus on technical tools, rather than humanistic questions, meant it remained on the margins of most humanities departments, housed in computer science labs or library special collections, rather than in English, history, or philosophy departments.

```mermaid timeline title 数字人文发展关键节点（1949-2020s） 1949 : 蒙特莱奥内完成《神曲》打孔机 concordance 项目 1946 : Busa 启动托马斯·阿奎那著作 concordance 项目 1960s :