Wednesday, October 12, 2005

Tufte - Visual Display/Quantitative Information (1983)

Excellent graphics are simple, clear pictures of numbers, Tufte argues in this "landmark" book, The Visual Display of Quantitative Information.  Basic graphical designs--"box plots, bar charts, historograms, and scatterplots" (124)--have in common principles of functional simplicity and clarity. Note the review comment attributed to the Boston Globe: "A visual Strunk and White." I read the first edition, and it's currently out in a second edition, so these notes should be so-understood.  They reflect the 1983 edition--the version that later needed an update for one reason or other. 

Tufte's book is filled with examples, and because the examples are illustrative--literally graphical--TVDOQI is good for browsing and for casual returns or future reference.  The principles guiding Tufte's project are fairly straightforward: "Graphics reveal data" (13).  Just how they reveal data best along with anti-examples or failures of statistical graphics to reveal data simply and clearly--these are the concerns guiding the book.  The pages are filled with successes and failures rel. to lie factors (57) and chart junk (107).  Tufte also looks at data-ink ratios; he argues that effective statistical graphics should make comparison easy to see.  They should tell the story of the data with design variables matching exactly to data variables and with minimal interference from decorative schemes and editorializing (59).  TVDOQI advocates tailoring tidy relationships between data and graphical representations of data.  Representation and reading, however, are not depicted as complicated, flexible or interpretive; they're strict activities understood as rigid (timeless, acontextual) universals and rigid psychological models of comprehension and visual experience.

Tufte's project is explicitly focused on quantitative data; he is quite direct about the primacy of clarity as the ultimate aim of the statistical graphics he's concerned with.  Because his project is bracketed in this way, it raises, for me, questions about a broader representative arena.  Why not The Visual Display of Qualitative Information(like city blocks with photographs at the beginning of  Sidewalk) or The Visual Display of Imaginative Information (mind-mapping and conceptual graphing)?  They're not quite the same type of scientistic data-sets Tufte looks at in TVDOQI, but his project is still useful for thinking about these other areas.

Tufte rationalizes several claims on a scarcity model of (re)presentational space (i.e. front pages of news papers or the confines of scientific journals); one of his formulas--the data-ink ratio (93)--falls right in line with a related theme: design efficiencies that presume normal, undifferentiated reading of statistical graphs.  Lasting and insightful as some of these codes might be, the critiques of moire effects (patterned shading) and pencil strokes for the drafting artist apply more roundly to the historical moment--how many turns of the straight-edge should a rug plot (135) require...and so on.

I had a few thoughts about small multiples (170)--one of the varieties of statistical graphics I hadn't thought much about before.  And I should return to the few pages near the end where Tufte articulates an integrative view of words, numbers and pictures (181).  Bits from the brief section on the integrative view seem to conflict with the consistent lines of data-as-truth and graphs-as-clear.  The example of a page from Leonardo's notebook is interesting, too. But all in all, Tufte, widely associated as he is with data visualization, tips heavily into constraint and reduction with TVDOQI

"The design of statistical graphics is a universal matter--like mathematics--and is not tied to the unique features of a particular language." (2)
"Graphics reveal data" (13).
"The distinguished graphic successfully organizes a large collection of numbers, makes comparisons between different parts of the data, and tells a story" (30).
"But no information, no sense of discovery, no wonder, no substance is generated by chartjunk" (121).

Bookmark and Share Posted by at October 12, 2005 7:30 AM to Reading Notes