spence@utstat.uucp (Ian Spence) (11/14/88)
Next Meeting: Wednesday, November 16 at 2.00 p.m. in Sidney Smith R.570
Topic: Graphs and psychophysics
Speaker: Ian Spence
Synopsis: Tufte, Macdonald-Ross, Cleveland and McGill, and others,
recommend that we do not construct graphs with graph
elements that have more dimensions than the data. The use
of boxes and cylinders, as well as volumes of other kinds,
is frowned upon, as is the use of graph elements that
require the observer to judge areas.
Such opinions are loosely based on results in traditional
psychophysics, a branch of psychology that seeks to
investigate the relation between subjective experience
and the physical magnitude of stimuli. It is well known
that judged length is proportional to physical length, but
that judged areas or volumes are generally underestimated,
with a power function providing a better description than
a straight line.
Until now, no one has determined the exponents of the
psychophysical function for common graphical elements, such
as lines, bars, pie and disk slices, boxes and cylinders.
An experiment that does this will be presented and
discussed, and a definitive answer to Tufte's question,
"Should we forego the 4000 lb chicken", will be given.