80 TEACHERs COLLEGE COLUmbIA UNIvERsITy
Developing a Cognitive model of Observational Drawing
of, or able to fully articulate, their thought processes
(or that of other artists). The language used to
describe such skills may be ambiguous or inconsis-
tent. A cognitive model may therefore contribute to
articulating drawing skills, and inform teaching and
assessment practices.
It could be argued that any written account of
drawing skill will be incomplete; teaching also occurs
through demonstration, non-verbal communica-
tion and feedback, in the context of teacher-student
relationships. Yet there remains a need to describe
drawing; those to whom teaching practices and cur-
riculum choices must be justified require concrete
reasoning and written material (for example, learn-
ing outcomes and level descriptors). My recent paper
(2011) explores these issues in more detail.
Developing a model of drawing and cognition
The study uses an observational methodology,
in a grounded theory framework, to study five art-
ists. Two members of the royal society of portrait
painters, and three PhD students. Footage of obser-
vational self-portrait drawing was recorded, and
concurrent and retrospective verbal reports were
elicited using methods outlined by Ericcson and
Simon (1993)^1. The resulting rich data were ana-
lyzed using a grounded theory approach.^2
A grounded theory approach to data analysis.
Rather than pre-determine a coding scheme
based on a specific hypothesis, the schemes for ana-
lysing the transcripts and video data were generated
using grounded theory methods, which allow for an
exploratory inductive approach.
In this approach, the coding scheme “emerges”
from the data itself. Through successive iterations,
the coding scheme is developed openly. Each time
a new code or category is needed, it is added. If
new data cannot be assimilated, the scheme is
modified. This process is repeated until the coding
scheme reaches “saturation point,” that is, new data
can be assimilated without further amendments. At
this point, all transcripts are coded again, using the
final scheme. (This process is outlined in Charmaz,
2006: 42.)
While it was originally intended that the cod-
ing scheme would be developed with the purpose
of describing and comparing observed behaviour, it
became clear that the scheme itself was becoming a
significant product of the research.
Development of the coding scheme.
The process of developing a coding scheme from
the concurrent report transcripts distinguished two
types of cognition. Many of the statements were
easily categorized according to their role in problem
solving strategies; goal setting, evaluation and deci-
sion making; or they revealed meta-cognitive pro-
cesses such as rationalisation. For example:
“I’m still right in the middle of the eye. Just try-
ing to coax it into something that I like.” (Concur-
rent report, AC)
Here the artist is monitoring his progress against
a subgoal: to redraw the eye in a satisfactory way.
However many statements only demonstrated evi-
dence of apprehending, for example:
“The glasses. There’s the bridge over the nose.
The two side pieces that are more obvious than the
glass, they help also describe the angle of the head.”
(Concurrent report, DC)
While these types of statements are the result
of underlying thought processes, such as decisions
about what to include in the drawing and routines
for locating salient features, it is nevertheless pos-
sible to categorize them at face value.
In some instances it was clear that the features
apprehended were used to inform evaluative pro-
cesses, as in the first example, while in others the
artists were simply looking and drawing. Initially
it seemed that the problem could be solved with an
additional category—“input” in which the artist was
simply apprehending features to be drawn. Within
this category the various types of features could be
categorised and nested as a further level of coding.
In this way, data about what the artists reported
attending to would be included. However, this nest-
ing was problematic, as while not all statements
mentioned features, all types of statements some-
times mentioned features, often more than one.
This created the need for every type of statement to
include this nested level.
A solution presented itself, which was simply to
parse and code twice using separate columns. For
example, AR makes a decision to change a portion
of the drawing:
“Ok. So that’s the sort of compositional size that
I want the head to fill. That’s the sort of space I’m
gonna use on the page. Ok. The problem there is I’ve
just gone too long. So I just need to reassess the shape
of the nose, the chin and the mouth in relation to the
eyes, so it’s that sort of triangle bit (gestures) in the
centre of the head.” (concurrent report, A R)