Pixels or Perish
The art of scientific illustration will have to adapt to the new age of online publishing
Scientists make pictures for many purposes. Doodles and sketches in a lab notebook might serve a strictly private function; many graphs and charts are created in a process of exploratory data analysis, and are soon discarded. Here I want to focus on more formal illustrations—those destined for publication, perhaps in a journal or an American Scientist article, perhaps in a textbook or on an educational website. And because my interests are computational, I’m going to emphasize quantitative graphics.
Any account of publication-quality computer graphics has to begin with PostScript, the “page description language” developed in the 1970s and 1980s by John Warnock and Charles Geschke, the founders of Adobe Systems. Warnock and Geschke are computer scientists, but they worked closely with graphic artists, typographers and the printing trade, and the language reflects this influence.
PostScript is primarily a language for “vector” graphics, where objects are constructed from geometric lines and curves, rather than “raster” graphics, where an image is a rectangular array of discrete pixels. PostScript operators with names such as moveto, lineto and curveto construct a path in a two-dimensional coordinate system of almost unlimited precision, so that the geometry of the drawing is independent of the resolution of the output device. Paths can be built from straight line segments or from curves called Bézier splines, defined by cubic equations. The operators stroke and fill can then be applied to create a visible graphic object. Some aspects of the language seem almost comically fastidious, such as the elaborate specifications of beveled, mitered or rounded joints between stroked lines; but it turns out such fussiness makes a real contribution to the visual quality of the finished artwork.
PostScript has another distinctive property: It is not just a notation for describing drawings but a complete programming language, with features such as conditional expressions, iteration and named procedures. In this way PostScript blurs the distinction between drawing a picture and writing a program.
An illustration published in American Scientist in 1990 offers an example. Robert V. Levine of California State University, Fresno, had written an article on “The Pace of Life,” measuring quantities such as walking and talking speed in 36 cities. As an aid to understanding this multivariable data, I experimented with a visualization technique invented by Herman Chernoff of Stanford University. The illustration mapped Levine’s measurements to various features of a cartoon face. The PostScript file that generated this figure did not specify the coordinates of the various lines, arcs and ellipses in each of the 36 faces; instead, it had a single face-drawing procedure, which was invoked 36 times on 36 rows of raw data. Thus the illustration didn’t exist, even as an internal data structure, until the program was run.