62 TEACHERs COLLEGE COLUmbIA UNIvERsITy
A multi-stage Attention Hypothesis of Drawing Ability
dently rated each drawing on overall accuracy using
an 8-point scale. Unsurprisingly, artists’ drawings
were rated as more accurate than those produced by
non-artists (see Figure 1 for images of the 3 highest-
rated and the 3 lowest-rated drawings).
Selection of visual information
As our participants created drawings, a major
problem they faced was the moment-to- moment
selection of what information from the model to
attend to and render—a ubiquitous problem for
anyone drawing from observation. Individuals who
draw from observation routinely look back and
forth from the model to the drawing because only
a small amount of information from the model can
be perceived and drawn at any given moment (Tch-
alenko, 2009). Thus, observational drawing behav-
ior involves a continual decision making process by
which the drawer must decide what information to
select to include in the drawing.
There is strong reason to believe that the par-
ticular information that is selected to be included
in a drawing will influence the perceived accuracy
of the drawing. Not all visual information is equally
important in supporting the recognition of that
object. For instance, Biederman (1987) reported an
experiment where he showed subjects line draw-
ings of different objects; in some, many line mid-
segments had been deleted; in others, vertices that
connected lines together were deleted. Participants
were significantly impaired in their ability to recog-
nize objects when the vertices were deleted, but not
when mid-segments were deleted. This suggests that
some forms of visual information, like vertices, are
much more important in supporting object recog-
nition than other forms, like line mid-segments. If
this is the case, then the type of visual information
that is selected to be included versus excluded in an
observational drawing should impact judgments of
the drawing as a recognizable representation of the
model stimulus. Further, since trained visual artists
produce more accurate drawings than non-artists,
skilled artists may be more sensitive to, and thus
are more likely to select and include in their render-
ings, the visual information that best contributes to
object recognition, compared to non-artists.
This hypothesis was tested in our experiment by
administering to our participants a limited line trac-
ing task (derived from Kozbelt, et al., 2010). This
task emphasized participants’ ability to select the
most important information to include in a depic-
tion. We provided participants with a grayscale
photograph of an elephant that was placed inside a
clear plastic folder. We asked participants to create
a tracing of the elephant on the clear folder. Partici-
pants did not use pencil or marker to create the trac-
ings, but rather, were given 30 short pieces of dark
brown tape with which to trace the photograph.
Each segment measured 2cm × 2mm. A piece of
white 8.5 × 11-in. paper was also available, so thatFigure 1. Examples of the free-hand drawings made by participants based on a photographic Model of an
octopus. The top row presents the top 3 rated drawings and the bottom row presents the bottom 3 rated
drawings.