SO40CH05-MoodyHealy ARI 4 July 2014 13:29
have tended to be generalizations of the scatter-
plot or barplot, either in the direction of seeing
more data or seeing the output of models. The
former looks for ways to increase the volume of
data visible, the number of variables displayed
within a panel, or the number of panels dis-
played within a plot. The latter looks for ways
to see results of models—point estimates, con-
fidence ranges, predicted probabilities, and so
on. Tufte (1983, p. 177) acknowledges that a
tour de force such as Minard’s “can be described
and admired, but there are no compositional
principles on how to create that one wonderful
graphic in a million.” The best one can do for
“more routine, workaday designs” is to suggest
some guidelines such as “have a properly cho-
sen format and design,” “use words, numbers,
and drawing together,” “display an accessible
complexity of detail,” and “avoid content-free
decoration, including chartjunk” (p. 177).
Among this set of general goals are some
specific details that can be employed to good
use across applications. This includes extensive
use of layering and separation, for example,
building on the insights of good cartography.
Judicious use of stroke weight and color allows
one to layer multiple meanings on a single
visual plane. The ability to successfully pull
off such effects depends on use of the smallest
effective difference—lighter lines, smaller color
variations, and simpler textures. It has long
been a complaint of chart designers that accom-
plishing this often means working very much
against the (highly detailed, drop-shadowed,
rich, Corinthian leather) grain of t he default
settings in spreadsheet or other chart-making
applications. Comparison and evaluation are
often enhanced by the use of many small
multiples—plots that repeatedly display some
reference variable or relationship (e.g., gross
domestic product versus health care costs over
time) and iterate across some other variable of
interest (e.g., country) in an ordered fashion
(see also Bertin 1967 [2010], pp. 217–45). The
use of such multiples highlights the notion of
parallelism that allows a reader to carefully
compare across instances of similar-but-
crucially-different items. Combined, these fea-
tures facilitate a simultaneous micro and macro
reading where key points are clearly communi-
cated at the surface, but deeper meaning is ob-
tained through careful review and exploration.
A common complaint about Tufte’s work
is that there are so few direct instructions.
Busy cooks want a cookbook, not a picture
of a fantastic meal. The tendency for the
codification of data visualization to vacillate
between overly abstract maxims and overly
specific examples is characteristic of any craft
where a practical sense of how to proceed—a
taste or feeling for the right choice—matters
for successful execution. A long-standing and
plausible response to the problem is to have the
designer make many of the judicious choices in
advance and then embed them for users in the
default settings of graphics applications. Given
that graphical software aimed at regular users
has been around for several decades now, how-
ever, these efforts have proven less successful
than initially hoped. In the foreword to the
new edition of Semiology of Graphics, Howard
Wainer (2010, p. xi) reflects on the hope he
and others once felt that easy-to-use graphical
tools and software would lead to better general
practice by way of smarter defaults. But, he
argues, this has not happened. In the end, high-
quality graphical presentation requires crafting
a deliberately designed message rather than
accepting the pre-established setting. Recent
theoretical work explicitly recognizes the limits
of relying on defaults. Following Wilkinson in
implementing ggplot’s “grammar of graphics”
for R, Wickham (2010, p. 3) notes that the
analogy to grammar is useful because although
“[a] good grammar will allow us to gain insight
into the composition of complicated graphics,
and reveal unexpected connections between
seemingly different graphics[,] ...there will
still be many grammatically correct but non-
sensical graphics. ... [G]ood grammar is just
the first step in creating a good sentence.”
If software defaults cannot enforce the ele-
ments of good taste, the next best—or maybe
better—thing is a means to easily expose the
mechanics of good practice. One of the most
positive developments in statistical software
110 Healy
·
Moody
Annu. Rev. Sociol. 2014.40:105-128. Downloaded from www.annualreviews.org
Access provided by Duke University on 08/09/17. For personal use only.