It’s been 50 years since Jacques Bertin’s Sémiologie Graphique was published. Bertin looms large in history of both data visualization and cartography. Before we had textbooks on data visualization by Munzner (Visualization Analysis and Design), Ware (Information Visualization: Perception for Design) or even Spence (Information Visualization: Design for Interaction), Bertin provided the theoretical foundation that much of visualization relies on today. Sémiologie Graphique structures the design space of visualization with the now familiar concepts of marks (point, line, area); visual attributes (only six in his version); type of perception (quantitative, ordered, categoric and associative); and layout. Beyond these, Bertin also considered many other aspects such as spatial separation (to form small multiples) ordifferent use cases of visualization including communication, analysis, and inventory.
However, only one aspect of Bertin’s work never made the translation from the French original to English: typography! Strangely, out of 450 pages+ only 4 pages on typography were not translated. In these 4 pages Bertin discusses the importance of the literal information represented by text. He notes that text is often the only encoding commonly accessible to both the textual/verbal system of encoding information; and the visualization system of encoding data.
Furthermore, Bertin points out that text is not selective. In other words, text is not preattentive, meaning that patterns do not automatically pop-out. If I ask you to find the word “six” occurring in the first paragraph, you need to linearly scan through the text – the benefits of visualization do not occur. Typographers would agree: they put significant effort in making text appear uniform with no visual anomalies to standout, carefully tweaking letterforms and kerning pairs to achieve this effect.
However, typographers also understand the need to make some words visually pop-out from surrounding text and therefore provide forms of emphasis such as italics and bold. Bertin also points out that these forms of emphasis are available and discusses them in the context of the technology of his time: pencil, pen, professional printers (which would have used phototypesetting in the late 60’s), and dry-transfer lettering (e.g. Letraset). And he nicely itemizes attributes of typography, available on page 415 of the French edition of Sémiologie Graphique:
So, why was Bertin wildly successful, but his commentary on typography so minimal that it was dropped from the English translation? Good question!
One answer is that even though Bertin indicates the potential of text to indicate data beyond literal text; he says the incremental addition of text only helps low-level elementary reading, not the higher level of visual perception of patterns. However, Bertin is writing in the late 1960’s: 15-20 years before Edward Tufte popularizes the notion that a visualization can be read at many different levels depending on the task, which Tufte calls micro/macro reading (Tufte: Envisioning Information).
Another answer is that even though Bertin acknowledges typographic attributes such as individual letter forms, typeface, width, spacing, size, weight, case and italic – he doesn’t provide any examples of the use of these attributes. On the otherhand, he provides hundreds if not thousands of examples of the other six visual attributes (size, orientation, hue, brightness, texture, shape), making sure that his core concepts are well explained and illustrated. A parallel can be seen in open-source visualization libraries: there were many different open-source visualization alternatives in the early 2010’s. However, Mike Bostock not only provided a well organized library with D3.js, he also provided a lot of compelling examples of visualizations implemented in D3 with source code. Mike made it far easier to adapt and extend D3’s model by starting with examples rather than requiring the extra effort to learn some other library and then figuring out how to create those examples.
There are other possible reasons, but the unfortunately reality is that Bertin’s typographic insights were side-stepped and never exposed to the English language research community. Bertin also wrote a follow-on article (Classification typographique : Voulez-vous jouer avec mon A doi : 10.3406/colan.1980.1369) specifically on the visual attributes of type in 1980 – but again, no examples and no translation (it does provide a better organization of the typographic attributes).
In Sémiologie Graphique, Bertin made 100 different visualizations of a dataset indicating three major occupations across 90 departments in France. None use typographic attributes (although a few use simple plain labels). I decided to make one typographic example – here’s Bertin’s ternary plot (p. 115) where the bubbles have been replaced with text, sized in proportion to population, colored based on occupation proportions. You can choose to focus on the macro patterns (e.g. most districts have an agricultural bias, or most of the agricultural districts tend to have smaller populations); or you can choose to focus on the micro details, e.g. district P (Paris) has the largest population and no agriculture; or, district 32 is the district with the highest proportion of people employed in agriculture. (If you want more non-Bertin examples, see many of my recent postings regarding typographic visualizations).
According to Google Scholar, there are 6004 citations for Bertin’s Sémiologie Graphique (across the French, German and English editions); while there is only ONE citation for Classification typographique : Voulez-vous jouer avec mon A. Perhaps this short article and references will help Bertin’s ideas of typography get more recognition and citations in future visualization and cartography research work.