aus der Introduction of the Busa Award Lecture bei der "Digital Humanities 2007" Konferenz
Urbana, 5 June 2007

C. M. Sperberg-McQueen


In the Foreword to the volume on Book VI of the Aeneid, there appears a sentence which seems to sum up, programmatically, both the capabilities and the potential limits of digital processing:

Electronic data processing can be put into service whenever data of any kind — notably including texts — must be processed according to rules which are unambiguously formulatable and completely formalizable.

What more compact formulation could we find of the fundamental program of our field? And what more matter of fact reminder that this is a description of those places where computers can successfully be deployed, without any suggestion of belief that they can be deployed absolutely everywhere.


In due course, these efforts produced a suite of programs for scholarly work with text, which at some point acquired the name Tustep, the Tübingen System of Text Processing tools.

Tustep embodied a number of important ideas:

  • completeness: It can be used for all parts of a project's normal work. There is an editor for data capture and revision, there are copy commands and tape utilities for archiving and moving data, there are a variety of general and specialized proceessing tools for manipulating documents, for sorting things, for extracting relevant items from lists, for laying documents out on the page, for photocomposition of the resulting pages, and so on.)
  • verifiability: Every action undertaken with Tustep will be logged, unless you take very active steps to avoid having it logged.
  • stability: Since projects may live for years or decades, the stability of the program and of Tustep files is critically important.
  • consistency: Years before anyone outside of Bell Labs had heard of Unix, Tustep adopted the principle that every tool would have one primary input and one primary output, and that the output of every tool would be usable as the input to any other tool. In practice, this means that the primary input and the primary output of each tool use the Tustep file format.

This principle of ensuring that other Tustep programs can read the primary output of any Tustep program is consistently implemented, even in cases where one might have expected a different choice. When I learned that the typesetting program of Tustep also produces a Tustep file as its output — the PDF or photocomposer file is, formally speaking, a side effect — I admired the consistent application of the design principle but privately thought that it bordered on the academic. Since for practical purposes the main output of a typesetting program is typeset pages, producing a Tustep file seems likely to be an anticlimax. What useful output can it produce? Perhaps just a copy of its input?

In fact, typesetting programs do produce information: before you run them, you don't know where the page and line breaks of the typeset pages will fall; afterwards, you do. So the primary output of the Tustep Satz program is a Tustep file showing the page and line breaks of the typeset version. This is why the production of a back-of-the-book index, in a project using Tustep, is mostly a routine matter of processing, rather than the eight-week crisis experienced by some editorial projects I have visited, who receive page proofs back from their publishers and must then hire every available graduate student to spend weeks translating the preliminary index from the working preliminary page numbering to the final page numbering of the volume. Consistent design decisions can have dramatic practical advantages.

The most important idea of Tustep, though, is that it is the responsibility of the software to serve the needs of scholarship, and not vice versa, and that the responsibility of the scholar is to respect the significant particularities of the material and the demands of his discipline (not any standards of practice imposed from outside, and least of all any limitations imposed by the software.)

Tustep developed over thirty years of listening to the needs of scholarship, consulting with projects and adding to Tustep the functionality they needed to enable them to do their work. Hundreds of editions have been prepared with it, some all the way from beginning to end, from data capture through typeset pages, others just translated into Tustep for the typesetting — apparatus criticus is not easy to set!

If we are to take responsibility, as humanists, for our use of machines, then it is necessarily now a part of humanities scholarship to understand and develop ways to make machines adapt to the requirements of our work, and (while remaining open to the exploitation of new and unforeseen opportunities) to resist the temptation to adjust our practices to suit the convenience of the machine. ...