The Computational Turn in the Humanities

The year 1964 marked an unlikely convergence of intellectual ambition and computational possibility. In that year, Father Roberto Busa—a Jesuit priest and Thomistic scholar—completed a twenty-nine-volume index of the works of Thomas Aquinas, a project that had consumed nearly three decades of meticulous labor. What made this achievement remarkable was not merely its scholarly significance, but the method by which it was accomplished: Busa had collaborated with IBM to process his index on punch-card computers, creating what is widely recognized as the first major humanities computing project. This moment, when a medieval philosopher's accumulated wisdom was rendered into machine-readable form, presaged a transformation that would reshape the landscape of humanistic inquiry over the subsequent six decades. The computational turn in the humanities did not arrive as a sudden revolution but emerged through a series of technological innovations, institutional adaptations, and theoretical reorientations that collectively redefined what it meant to study, interpret, and understand cultural materials in the digital age.

The intellectual trajectory from Busa's concordance to contemporary digital humanities encompasses far more than the adoption of new tools; it represents a fundamental reconceptualization of the relationship between humanists and their source materials. Where traditional scholarship prized the close, sustained engagement with individual texts—reading and rereading, marginal annotation, the slow accumulation of interpretive insight—computational methods introduced the possibility of analyzing thousands or millions of texts simultaneously. This shift from the particular to the general, from the intensive to the extensive, provoked sustained debate within the humanities about the nature of scholarly knowledge, the validity of computational evidence, and the proper role of algorithms in interpretation. Understanding this transformation requires attending to both the technical infrastructures that enabled new forms of scholarship and the theoretical frameworks that sought to make sense of what such scholarship could accomplish.

The Emergence of Text Encoding: From Concordances to Standards

The earliest computational work in the humanities was predominantly concerned with the creation of indexes, concordances, and word-frequency lists—tools that automated the most laborious aspects of philological research. The KWIC (Keyword in Context) concordance, developed in the early 1960s, represented a significant advance over manual indexing, allowing researchers to retrieve every occurrence of a word within a corpus along with its surrounding textual environment. These early systems operated on relatively simple principles: texts were converted into machine-readable form through keypunching or optical character recognition, then processed by programs that sorted and organized lexical data according to predetermined parameters. The limitations of these early systems—including their dependence on proprietary formats, their inability to capture textual structure beyond the level of the word, and their general inaccessibility to researchers without technical expertise—would drive the development of more sophisticated approaches throughout the 1970s and 1980s.

The breakthrough came with the development of markup languages designed specifically for the representation of textual structure. The Generalized Markup Language (GML), created by Charles Goldfarb, Edward Mosher, and Raymond Lorie at IBM in the late 1960s, introduced the foundational insight that textual elements could be identified through descriptive tags embedded within the document itself. Rather than treating a text as an undifferentiated string of characters, GML allowed markup to indicate structural features such as paragraphs, headings, emphasis, and quotations. This concept of "descriptive" rather than "procedural" markup—identifying what something is rather than how it should be processed—proved enormously influential and would eventually culminate in the Standard Generalized Markup Language (SGML), published as an international standard in 1986. SGML provided a formal syntax for defining markup languages through Document Type Definitions (DTDs), enabling the creation of standardized formats for specific document types.

The Text Encoding Initiative (TEI), launched in 1987 with funding from the National Endowment for the Humanities, the Council of Library Resources, and the Association for Computers and the Humanities, represented the most ambitious application of SGML principles to humanities scholarship. The TEI Guidelines, first published in 1994 and now in their third edition, established a comprehensive vocabulary for describing the structural, semantic, and renderable features of textual documents. The TEI's modular architecture allowed encoding projects to select and customize elements appropriate to their materials while maintaining interoperability with other TEI-encoded resources. This commitment to interoperability—ensuring that encoded texts could be exchanged, combined, and analyzed across institutional and disciplinary boundaries—became a defining characteristic of the digital humanities infrastructure.

The TEI's influence extended far beyond its immediate utility as a encoding standard; it embodied a particular philosophical stance toward textual representation that continues to shape the field. The distinction between transcription and interpretation, between recording what can be observed in a document and encoding what one believes it means, became a central organizing principle for TEI practice. The guidelines explicitly accommodated both levels of description through separate modules for simple transcription and for more complex interpretive encoding. This theoretical sophistication—the recognition that encoding involves countless decisions about significance, relevance, and meaning—distinguished the TEI from simpler markup approaches and established it as a framework for humanistic reflection on the nature of textual representation itself.

Consider the encoding of a manuscript page containing a medieval poem. A minimal transcription would record the sequence of characters, including any abbreviations or damaged areas indicated through editorial conventions. A structural encoding would identify verse boundaries, stanza divisions, and any rubrication or marginal annotations present in the manuscript. An interpretive encoding might identify the author, date, and provenance of the text, annotate names and allusions, and encode the critical apparatus recording variant readings. The TEI's modular structure allowed encoding projects to begin with minimal transcription and progressively add interpretive layers as scholarship developed, without invalidating earlier work. This extensibility proved crucial for long-term projects that might span decades and involve multiple generations of scholars with evolving theoretical commitments.

Structural Transformation: XML and the Digital Documentary Ecosystem

The transition from SGML to XML (Extensible Markup Language) in the late 1990s marked a significant inflection point in the development of digital humanities infrastructure. While XML retained SGML's fundamental approach to markup—hierarchical structures, descriptive tags, separation of content from presentation—it simplified the syntax in ways that facilitated adoption beyond the specialized community of SGML practitioners. The World Wide Web Consortium's promotion of XML as the foundation for web services, data exchange, and document processing created an ecosystem in which XML skills became valuable across academic and industry contexts. Digital humanities projects increasingly adopted XML as their primary format, benefiting from a wealth of tools, libraries, and documentation developed for the broader XML ecosystem.

The consequences of this shift were both practical a