Vous êtes sur la page 1sur 17

Uncertainty and sketching behaviour

Stephen A. R. Scrivener, Visual Information Design Research Centre,


Coventry University, Priory Street, Coventry, CV1 5FB, UK
Linden J. Ball, Institute of Behavioural Sciences, University of Derby,
Mickleover, Derby, DE3 5GX, UK
Winger Tseng, Visual Information Design Research Centre, Coventry
University, Priory Street, Coventry, CV1 5FB, UK

Evidence suggests that sketches of seen objects exhibit a structure


consistent with the volumetric parts of the recalled objects.
Occasionally, sketching structure does not match volumetric part
structure. Top-down cognitive factors, perception, or a combination of
both could trigger switching of drawing behaviour. From the evidence,
it is concluded that uncertainty is the primary factor triggering change
in drawing structure. Nevertheless, the data indicate that sketches can
engender uncertainty by revealing errors in recollection, through
recognition failures or by promoting reasoning about depicted objects,
and can facilitate the resolution of uncertainty by stimulating recall.
 2000 Elsevier Science Ltd. All rights reserved

Keywords: conceptual design, design cognition, reasoning, uncertainty,


drawings

1 Herbert, D M ‘Study drawings

D
in architectural design: their esign seems inconceivable without drawings. Produced drawings
properties as a graphic medium’ are used at all stages of design for many different purposes.
Journal of Architectural Edu-
cation Vol 41 No 2 (1988) Designers often distinguish between freehand drawings produced
pp 26–38
2 Goldschmidt, G ‘The dialec- in the early and late stages of design1,2, regarding the former as private
tics of sketching’ Design Studies tools for thinking rather than public communication and presentation aids.
Vol 4 No 2 (1991) pp 123–143
3 Scrivener, S A R ‘The inter- Vagueness, incompleteness and ambiguity are among the properties of
active manipulation of unstruc-
sketches taken as indicative of this function1,3–5. According to Goldsch-
tured images’ International Jour-
nal of Man-Machine Studies Vol midt2, design thinking is a dialectic between different forms of ‘seeing’,
16 (1982) pp 301–313
4 Goel, V Sketches of Thought seeing-as and seeing-that, that draw on different cognitive processes and
MIT Press, Cambridge MA knowledge types. Translation between modes of seeing is thought to both
(1995)
5 Fish, J and Scrivener, SAR develop and stimulate new ideas through reinterpretation. Systematic stud-
‘Amplifying the minds eye:
sketching and visual cognition’
ies are beginning to clarify the role of idea sketches in this cyclic pro-
LEONARDO Vol 23 No 1 (1990) cess2,4,6–11. It may be debatable whether sketches, via perception, are a
pp 117–126
6 Goldschmidt, G ‘On visual driving force for discovery or simply aid creative cognition (cf. Purcell
design thinking: the vis kids of
and Gero9), but it is becoming clear that sketching and cognition are closely
architecture’ Design Studies Vol
15 (1995) pp 158–174 coupled. If these processes are closely coupled then cognitive processes
www.elsevier.com/locate/destud
0142-694X/00 $ - see front matter Design Studies 21 (2000) 465–481
PII: S0142-694X(00)00019-3 465
 2000 Elsevier Science Ltd All rights reserved Printed in Great Britain
7 Schon, D A and Wiggins, G
‘Kinds of seeing and their func-
and structures should be reflected in sketching structure, which indeed
tion in designing’ Design Studies seems the case12–15. We, too, are concerned with what sketching structure
Vol 13 No 2 (1992) pp 135–156
8 Suwa, M and Tversky, T reveals of cognitive processes and structures. We begin by reviewing data
‘What do architects and students
perceive in their design
supporting the hypothesis that drawings of recently seen objects are seg-
sketches? a protocol analysis’ mented according to geometric components. However, our principal inter-
Design Studies Vol 18 No 4
(1997) pp 385–403 est are those data that go against the hypothesis. In particular, what
9 Purcell, A T and Gero, J accounts for this unpredicted behaviour and what does it tell us about
‘Drawings and the design pro-
cess’ Design Studies Vol 19 No underlying cognitive processes?
4 (1998) pp 389–430
10 Suwa, M, Purcell, T and
Gero, J ‘Macroscopic analysis of
design processes based on a
scheme for coding designers’
1 Structure in sketching behaviour
cognitive actions’ Design Studies
Kavakli et al.15 conducted a study investigating whether sketching behav-
Vol 19 No 4 (1998) pp 455–484 iour reflects the geometric structure of the objects represented. Novice
11 Verstijnen, I M, Van
Leeuwen, C, Goldschmidt, G, designers undertook five sketching tasks. Three required them to observe
Hamel, R and Hennessey, JM
‘Sketching and creative dis-
and then draw from memory three chairs, called the Library, School and
covery’ Design Studies Vol 19 Café chairs (Figure 1). The fourth required them to design a chair from
No 4 (1998) pp 519–546
12 van Summers, P Drawing imagination and the last to overtrace their drawings.
and Cognition Cambridge Uni-
versity Press, Cambridge, MA
(1994)
13 Novick, L R and Tversky, The chairs were conceived as composed of volumetric components or parts
B ‘Cognitive constraints on (Figure 2). The participants’ depicted chairs defined the geometric structure
ordering operations: the case of
geometric analogies’ Journal of to be analysed, i.e. this was not a memory test. If a designer drew one of
Experimental Psychology: Gen-
eral Vol 116 (1987) pp 50–67
these parts completely, before moving on to draw the next part, they were
14 Taylor, H A and Tversky, B said to be drawing Part by Part. Consider Figure 5(c) where parts P5 and
‘Descriptions and depictions of
environments’ Memory and Cog- P6 were coded Non Part by Part as they are clearly represented incom-
nition Vol 20 (1992) pp 483–496
15 Kavakli, M, Scrivener,
pletely. Had part P5 in Figure 5(c) been represented completely, like P5
SAR and Ball, LJ ‘Structure in in Figure 2(a), before drawing commenced on another part then it would

Figure 1 The Library, School and Café chairs

466 Design Studies Vol 21 No 5 September 2000


Figure 2 Parts of chairs (a) Library chair; (b) School char; (c) Café chair

have been coded Part by Part. Often participants began with ‘stick’ draw-
ings that were then elaborated (cf. Figure 6, drawing D1). The legs of
drawing D1 in Figure 6 were coded as Part by Part because, although
lacking geometrical detail, the full extent of the parts is represented. Thus
the judgement between Part by Part and Non Part by Part drawing is not
made on the basis of geometrical detail but on whether or not the represen-
tation captures the fundamental geometry of the part.

A significant proportion of drawn parts, whether from memory, imagin-


ation or by overtracing, were produced Part by Part15. Overall, sketching
behaviour corresponded with the volumetrical part structure of the objects
depicted. Nevertheless, some parts were represented Non Part by Part and
we analyse these data to throw further light on when and why designers
switch to Non Part by Part drawing. As the memory task required parti-
cipants to recall an actual chair, distinguishing between drawn parts was
straightforward using the actual chair and/or their recalled chairs for refer-
ence. The design task was problematic because one had to decide whether
a drawn part was a version of an existing part or a newly invented one,
and hence these data are less reliable. Overtracing data are excluded since
none are produced in the absence of other drawings. Hence, we only con-
sider data emerging from the memory tasks in subsequent sections.

2 Non part by part drawing in the memory tasks


Table 1 presents the Part by Part and Non Part by Part drawing results for
the three memory tasks. Across participants and chairs the majority of parts
are drawn as complete representations of the volumetric components of
the recollected object. The data reveal sketching behaviour where typically
idea sketching behaviour’ a complete part of a remembered chair is drawn before the next complete
Design Studies Vol 19 No 4
(1998) pp 485–518 part, and so on. Nevertheless, 19.3% of parts are represented Non Part by

Uncertainty and sketching behaviour 467


Table 1 Part by Part and Non Part by Part drawing by task and participant

Library School Café Totals %

Participant Part by Non Part Part by Non Part Part by Non Part Part by Non Part Part by
Part by Part Part by Part Part by Part Part by Part Part

1 17 2 7 3 6 3 30 8 78.9
2 10 10 18 0 15 3 43 13 76.8
3 16 6 26 3 20 4 62 13 82.7
4 34 0 34 6 7 8 75 14 84.3
Total 77 18 85 12 48 18 210 48
%Part by 81.1 87.6 72.7 80.7 19.3
Part

Part and therefore occasionally drawing behaviour switches from Part by


Part to Non Part by Part, and back. Our goal is to examine possible causes
of this change in drawing behaviour.

In order to understand the determinants of changes in drawing behaviour


we assume that drawing reflects either high-level cognitive activity (e.g.
the manipulation of different knowledge representations), lower-level cog-
nitive activity (e.g. perceptually-driven recognition) or a mixture of both
types of activity (see Purcell and Gero9 for a relevant discussion of the
debate surrounding these possibilities). As such, we assume that changes
in drawing behaviour may be engendered by shifts in cognitive state arising
from: (i) top-down cognitive factors without any contribution from the
perception of drawing stimuli; (ii) top-down cognitive factors with a contri-
bution from the perception of drawing stimuli; (iii) purely perceptual fac-
tors, driven bottom-up by drawing stimuli.

3 Drawing stimuli as a determinant of cognitive shift


3.1 Non part by part drawing in session first parts, and
first parts
If perception of drawing stimuli is the only cause of Non Part by Part
production, then we should expect the first part drawn in each session, the
Session First Part, to be Part by Part. In fact, 2 of these 12 parts are drawn
Non Part by Part, suggesting that a cognitive factor is involved in Non
Part by Part production, making Explanation (iii) (see previous section)
unlikely.

At the onset of a session participants must reproduce the remembered chair


onto a virgin sheet. As the session progresses each chair part is drawn out

468 Design Studies Vol 21 No 5 September 2000


of memory. If pre-drawn parts determine those being produced for the first
time, then we might expect this to be reflected in changes in Non Part by
Part production. Of the 90 First parts produced 13.3% are drawn Non Part
by Part. While the percentage of Non Part by Part production in Session
First parts, 16.6%, is greater than that for other First parts, 12.8%, most
of the Non Part by Part production of First parts (75%) occurs in the first
drawing in each session. The Non Part by Part: Part by Part ratio in Session
First parts is similar to that for other First parts in first drawings (i.e. 0.166
vs 0.179) suggesting performance in first drawings is unaffected by the
presence of other drawn parts. In contrast, the First part Non Part by Part:
Part by Part ratio is markedly different between the first drawings and later
session drawings (i.e. 0.214 vs 0.076).

Summarising, when parts are first externalised from memory they are
occasionally produced Non Part by Part. Of all First parts, 56.6% are pro-
duced in the Session First Drawings, with Non Part by Part production
apparently unaffected by the presence of parts drawn earlier, suggesting
this behaviour is cognitively determined. However, the proportion of First
parts produced Non Part by Part decreases markedly in later drawings com-
pared to Session First Drawings. Since Session First Drawings production
does not appear to benefit from drawing stimuli it is unlikely that reduction
in Non Part by Part in later drawings is principally due to drawing stimuli.
Rather, this variability in Non Part by Part production of First parts pro-
vides further support for top-down cognitively determined switching
involving factors that diminish in intensity as the task progresses.

3.2 Non part by part over drawings


Only seven drawings are produced in Drawings 7–10, comprising only
8.5% of all parts (see Table 2), so subsequent discussion will focus on
Drawings 1–6. Table 2 shows that Non Part by Part production as a per-
centage of the total number of parts produced is relatively constant across

Table 2 Frequency of drawing number, parts and Non Part by Part percentages

Drawing 1 2 3 4 5 6 7 8 9 10
number

Drawing 12 12 10 7 6 5 2 2 2 1
frequency
Part 51 57 55 26 23 24 9 6 3 4
frequency
% Non 18 7 27 23 35 4 22 33 0 25
Part by
Part

Uncertainty and sketching behaviour 469


drawing number except for Drawings 2 and 6 where Non Part by Part
production drops markedly. The total number of parts in second drawings
is 57, 22 being First parts and 35 Repeat parts, of which four are Non Part
by Part parts. In Drawing 3, 14 of 44 Repeat parts are Non Part by Part.
Hence, the proportion of Non Part by Part production of Repeat parts
increases markedly in Drawing 3 (31.8%) as compared to Drawing 2
(11.4%). Overall, the proportion of Non Part by Part production increases
in Repeat parts compared to First parts (i.e. 22.4 vs 13.3%). Yet, Drawing
2 exhibits a lesser proportion of Non Part by Part First parts than Drawing
1 (i.e. 0 vs 17.6%), and a lesser proportion of Non Part by Part Repeat
parts than Drawing 3.

How can we explain the reduction in Non Part by Part production in Draw-
ing 2 vs Drawings 1 and 3? Thus far, the evidence suggests that changes
in drawing behaviour in First parts are determined by cognitive factors that
are less evident as the task progresses, hence the drop in Drawing 2. Let
us assume, however, that drawing stimuli only become a driving force as
parts are repeated. Since the effect is weaker in Drawing 2 compared to
Drawing 3, we might expect Non Part by Part production to increase with
drawings. However, this does not appear to be the case from Table 2, and
Non Part by Part production decreases in Drawing 6. If drawing stimuli
are the primary determinants of cognitive shift then their effects are com-
plex. Again this suggests the involvement of top-down cognitive factors
in the variable proportion of Non Part by Part production over the first
three drawings.

Therefore, we conclude that the change in Non Part by Part production


over drawings reflects cognitive shifts determined by top-down cognitive
factors or these interacting with drawing stimuli. Given this interpretation,
the data suggest these factors arise less in Drawing 2 than other drawings
(apart from Drawing 6). This is of interest as there is an important qualitat-
ive difference between Drawings 1 and 3. As we have seen Drawing 1 are
comprised entirely of First parts, i.e. parts first realised entirely from mem-
ory, whereas Drawing 3 is comprised largely of redrawn parts (80%), i.e.
parts for which versions already exist in drawing stimuli. Given this quali-
tative difference it may be that the cognitive factors influencing Non Part
by Part production are also different between Drawing 1 and Drawing 3,
and beyond. However, the question remains as to what are the cognitive
factors that trigger the cognitive shifts evidenced by Non Part by Part
production and whether these cognitive shifts are influenced at all by draw-
ing stimuli.

470 Design Studies Vol 21 No 5 September 2000


4 Cognitive synthesis and representational mode as
determinants of cognitive shift
4.1 Cognitive synthesis
Kavakli et al.15 suggest that Non Part by Part production might be related
to cognitive synthesis, i.e. forming mental wholes from remembered sub-
wholes that have been previously imaged16. Kavakli et al.15 argue that such
synthesis might require or be facilitated by functional knowledge of the
object, hence yielding Non Part by Part drawing. However, this does not
appear to be what happens. By the completion of Drawing 2, 81.1% of
first parts have been produced. Hence, Drawings 1 and 2 represent much
of the recalled chairs, suggesting that for this task objects are not mentally
synthesised in sub-wholes and then combined (i.e. this is not reflected in
the drawings).

4.2 Functional versus structural representation


Kavakli et al.15 argue that the predominance of Part by Part production
indicates that drawing production is driven by structural rather than func-
tional representations. They suggest that drawing behaviour might be tied
to these different kinds of representation, cognitive shift being a change
of one for the other (cf. Tversky and Hemenway17). However, each chair
is largely comprised of uni-functional components that do not enable a
distinction to be made between function or structure as Part by Part pro-
duction of these parts is consistent with either a functional or structural
representation. To explore this function/structure distinction further, we
focussed on the multi-functional components of the chairs. If these compo-
nents are produced more often Non Part by Part than Part by Part then this
would suggest a functional representation.

Table 3 shows the multi-functional parts as recalled by the participants.


Three out of the four designers failed to remember the back-leg compo-
nents of the Library chairs. In each case, separate back and leg components
were produced. In the case of the School chair, all designers recalled the
back-seat component and the double leg components. However, two of the
four failed to recall the double leg components accurately. Participant 1
recalled all four legs as being connected to a single cross member, while
Participant 2 recalled the double leg component as comprising a front and
16 Finke, R A Principles of back leg. All participants accurately recalled the arm-leg component of the
Mental Imagery MIT Press,
Cambridge, MA (1989) Café chair apart from Participant 3 who failed to recall that the chair had
17 Tversky, B and Hemen-
way, K ‘Objects parts and cate- arms, producing a multi-functional component comprising a back frame
gories’ Journal of Experimental and rear legs. Multi-functional parts proved more difficult to remember
Psychology: General Vol 113
(1984) pp 169–193 than uni-functional parts.

Uncertainty and sketching behaviour 471


Table 3 Multi-functional chair parts as recalled by participants (1 = Non Part by Part, 0 = Part by Part)

Library School Café

Participant 1 – Back-seat 1 Arm-leg 0


– Front legs 1
Back legs 1
Participant 2 – Back-seat 0 Arm-leg 0
– Front-back left 1
Front-back 1
right
Participant 3 Back-leg left 0 Back-seat 0 Back-leg 0
Back-leg right 0 Front legs 0
Back legs 0
Participant 4 – Back-seat 0 Arm-leg 0
– Front legs 0
Back legs 0

Table 3 shows which of these multi-functional parts were first produced


Non Part by Part (coded 0) or Part by Part (coded 1). Of these participant-
produced multi-functional parts, 27.7% were produced Non Part by Part,
which is a higher percentage than for all First parts (13.3%). Nevertheless,
72.3% of multi-functional parts are produced Part by Part supporting the
proposition that structure dominates function in drawing. Of course, these
parts are also geometrically more complex than uni-functional parts which
may also explain their poorer recall. However, it is unlikely that Non Part
by Part production of these parts is simply due to geometrical complexity
as multi-functional parts are often drawn in fragments terminating at func-
tional boundaries. Consequently, we cannot discount the possibility that
one effect of cognitive shift is selection between structural or functional
representations, reflected in Non Part by Part drawing.

5 Uncertainty as a determinant of cognitive shift


Above we have suggested that variation in Non Part by Part production
over drawings reflects cognitive shifts. So far the analysis does not support
mental synthesis as a determinant of cognitive shift, while change in rep-
resentation with cognitive shift cannot be discounted. In this section, we
undertake a qualitative analysis of the Non Part by Part production in all
first, second, third and subsequent drawings to throw further light on the
cognitive determinants of Non Part by Part production. We will argue that
uncertainty is a primary determinant of the cognitive shifts giving rise to
Non Part by Part production.

Human memory, whilst often highly robust, is also known to be fallible


and liable to omissions, distortions, abstractions, and constructive influ-

472 Design Studies Vol 21 No 5 September 2000


ences such as those deriving from prior expectancies or stereotypical
beliefs18. Some memory retrieval difficulties also appear to be associated
with feelings of uncertainty. For example, tip-of-the-tongue states and feel-
ing-of-knowing states are examples of situations where people believe they
possess relevant information in memory but experience difficulty in
accessing it. An interesting aspect of such memory-engendered uncertainty
about stored or partially-retrieved memories is that it may actually promote
strategic shifts in processing activity aimed at resolving the uncertainty.
This view has been supported by much evidence derived from studies of
strategy selection in memory and problem solving situations20,21. Generalis-
ing such ideas to design, Ball et al.22 have presented evidence that uncer-
tainties about current solution ideas cause major strategic switches in the
expert designer’s ongoing problem solving (e.g. changes from optimal bre-
adth-first design to novice-like depth-first or opportunistic design).
18 Baddeley, A D Human
Memory: Theory an Practice
Revised edition, Psychology
Here, we use the term uncertainty to refer to a cognitive state caused by
Press, Hove (1997) the lack of information in memory necessary to satisfy the needs of the
19 Cohen, G Memory in the
Real World Lawrence Erlbaum task, i.e. to enable a chair to be drawn to the complete satisfaction of the
Associates, Hove (1989)
20 Cavanaugh, J C ‘The place
participant. Recognition of this lack could be detected during visualisation
of awareness in memory devel- or during drawing. Additionally, data provided by perception may enable
opment across adulthood’ in L W
Poon, D C Rubin and B A Wil- detection of inadequacies in recollection. We suggest that these moments
son (eds) Everyday Cognition in of realisation are experienced as uncertainty because they impede the task
Adulthood and Later Life Cam-
bridge University Press, Cam- and, until they are resolved, what is to be done next is uncertain.
bridge, MA (1988)
21 Roberts, M J and Erdos, G
‘Strategy selection and metacog-
nition’ Educational Psychology
5.1 First drawings
Vol 13 (1993) pp 259–266
As noted earlier, Non Part by Part production occurs in two first drawings
22 Ball, L J, Evans, J St B T, (Figure 3). The general impression in both cases is that the participants
Dennis, I and Ormerod, T C
‘Problem-solving strategies and are uncertain about the components themselves and/or how they join
expertise in engineering design’ together. This is supported by the fact that both Participants 1 and 2 fail
Thinking and Reasoning Vol 3
(1997) pp 247–270 to remember correctly the structure of the legs. Although Figure 3(a) indi-

Figure 3 Non Part by Part


first drawings (a) Partici-
pant 1’s School chair; (b)
Participant 2’s Library
chair

Uncertainty and sketching behaviour 473


cates that Participant 1 has remembered correctly that the two front and
back legs are each created from single pieces of tubing that are then joined
together, when the drawing is overtraced the legs are shown in an arrange-
ment quite different from the actual chair. Similarly, although Participant
2 appears to remember that a component functions as both back support
and leg, when redrawn in Drawing 3 (see Figure 5(b)) the back support
and leg are shown as separate, therefore invented, components. Thus both
participants fail to correctly remember some aspects of the respective chairs
and must design their way out of this uncertainty, by determining the struc-
ture of components and their relationship to each other.

5.2 Second drawings


Non Part by Part production occurs in only one of the 12 second drawings,
i.e. Participant 4’s Café chair. In Drawing 1, the designer draws the com-
plete chair entirely Part by Part. Drawing 2 is also a complete drawing of
the chair although here it is seen in a rotated view showing the back (Figure
4). The designer partly draws the seat, followed by a partly drawn arm-
leg, left and right rear legs, and finally the back, which is the only part
drawn Part by Part. Comparing Figure 4 to Figure 1(c), it is clear that the
designer recalls the chair with a back-leg member that is not present in
the actual chair, i.e., the rear leg of the actual chair terminates at the seat.
Inspection of the near rear leg (Figure 4), suggests some uncertainty about
whether this invented component is a single multi-functional part or two
uni-functional components, i.e. leg and back support. Perhaps the designer
has a vague recollection that the back leg terminates at the seat that creates
doubts in his mind about the multi-functional part depicted in Drawing 1.
Drawing 3 (Figure 5(f)) supports the idea that this uncertainty exists as it
clearly presents an alternative in which the rear components are shown as
leg and back.

Figure 4 Non Part by Part


second drawings; Partici-
pant 4’s Café chair

474 Design Studies Vol 21 No 5 September 2000


Figure 5 Non Part by Part third drawings (a) Participant 1’s Library chair; (b) Participant 2’s Library chair; (c) Participant
3’s Library chair; (d) Participant 3’s School chair; (e) Participant 4’s School chair; (f) Participant 4’s Café chair

5.3 Third drawings


Non Part by Part drawing occurs in six of the ten third drawings (Figure
5). Considering Participant 1’s Library chair, Drawings 1 and 2 are ortho-
graphic front and side projections, whereas Drawing 3 is a perspective
drawing. Having first drawn the front near leg (Figure 5(a)), followed by
some construction lines, the designer then shows the depth of the seat at
the rear near leg before partially drawing the far front and far rear corners
of the seat. Here, then, Non Part by Part production seems to reflect the
needs of accurately reproducing the conjunction of parts. Like Participant
2, this designer recalls the back-leg component incorrectly as two compo-
nents. Given that this is an incorrect recollection, how the parts join at the
seat must also be undefined, or potentially in conflict with any recollection
of their actual alignment. Thus, although this Non Part by Part production
may be qualitatively different to that occurring in Drawing 1 described
above, as the conjunction of seat, legs and back supports has been pre-

Uncertainty and sketching behaviour 475


viously defined in Drawings 1 and 2, uncertainty may still play a part in
Non Part by Part production.

As noted above, Participant 2 appears confused as to whether or not the


back-leg component is one or two elements. In Drawing 3 (Figure 5(b)),
the designer first seems to settle on the correct recollection of a single
back-leg component, only to then redraw it as separate back and leg
components. Again, uncertainty appears to figure strongly in Non Part by
Part production. Later we shall return to consider what might cause a cor-
rect recollection to be discarded for an invention.

Participant 3’s first and second perspective drawings of the Library chair
are of single parts, the first showing the back-leg component, the second
the back. In Drawing 3 (Figure 5(c)), also in perspective, the designer first
draws the seat and near and far front legs completely before sketching only
the leg portions of the back-leg components, stopping in each case just
above the conjunction of seat and leg. Here the designer seems quite certain
about the back-leg component. However, the fact that he stops drawing at
the conjunction suggests some uncertainty, as otherwise why not complete
the entire legs? On the other hand, efficiency could account for Non Part
by Part production here.

When drawing the School chair (Figure 5(d)), Participant 3 draws the seat
completely, the hand hole in the seat and the leg-to-seat attachment bolts.
The designer then draws the back leg and then the front legs before adding
some detail to the seat. Given that it is a perspective drawing, at first only
the visible parts of the front and back legs are shown. However, having
embellished the seat the designer returns and draws in the occluded frag-
ments of the legs, correctly recalled in Drawing 1. Clearly, this disfigures
the chair’s mimetic realism. One reason for drawing in the occlusions
might be that the participant recognises that the bolts are inaccurately pos-
itioned with respect to their veridical mental model of the chair. In this
case Non Part by Part production would be in the service of accurate
description and is unlikely to register uncertainty. Alternatively, the partici-
pant might have realised that the bolts in the drawing (which in the actual
chair function to connect the seat to the legs) serve no proper function.
This realisation of a functional anomaly would produce uncertainty about
their conception of the chair.

When drawing an orthogonal projection of the School chair (Figure 5(e)),


Participant 4 draws the seat of the seat-back component and then the legs
before returning to complete the back part of the seat-back. This can either
be regarded simply as an artefact of the method of projection, or as a

476 Design Studies Vol 21 No 5 September 2000


function of depicting the conjunction of parts. Uncertainty does not seem
to figure here. Finally, when producing a rear perspective view of the Café
chair (Figure 5(f)), Participant 4 draws the seat, arm-leg, near and far back
legs incompletely. Here, again the designer seems to be exploring the
possibility that that which has been recalled as a back-leg component may
comprise several components. The rear view projection enables attention
to be focused on a local arrangement of parts that is not clearly visible in
the earlier drawings. This change of understanding is also evident in the
overtraced drawings where Drawing 2 is clearly traced as having a back-
leg component and Drawing 3 separate back and leg components.

We have noted above that Non Part by Part production in Drawings 1 and
2 appears to be connected with uncertainty. This may be uncertainty about
the structure of components and/or their conjunction. We are drawn to
conclude that Non Part by Part production in Drawing 1 is a feature of
the resolution of this uncertainty. A number of explanations for Non Part
by Part production in Drawing 3 seem plausible (e.g. uncertainty, accuracy
and mode of representation). Nevertheless, uncertainty is again prominent
in our interpretations. Some Drawing 3 Non Part by Part production occurs
in the absence of component and conjunction information in earlier draw-
ings, but in most cases information about the parts drawn Non Part by Part
already exists. Since, typically, both parts and their conjunction have
already been outlined with some commitment, the impression is that Non
Part by Part production is related to uncertainty about the precise form and
arrangement of parts. Analysis of the 10 remaining drawings (8 of which
reproduce earlier drawings) that exhibit Non Part by Part production in
Drawings 4–10 also reveals both instances of uncertainty (i.e. general
and particular).

5.4 Discussion
We are now in a position to interpret the variations in Non Part by Part
production between Drawings 1 and 3. We have already concluded that
this is unlikely to be an effect of drawing stimuli. The qualitative analysis
of drawings above draws us to the conclusion that uncertainty is a primary
factor determining the cognitive shifts responsible for Non Part by Part
production. When a previously perceived object is recalled from memory
it is recalled part by volumetric part (e.g. Drawings 1 and 2). A cognitive
shift may be necessary to resolve uncertainty about the components them-
selves or how they are combined leading to Non Part by Part production
(e.g. Drawing 1). Once initial uncertainty is resolved, the remaining parts
can be extracted from memory Part by Part (e.g. hence the fall in Non Part
by Part production in Drawing 2). Having extracted First parts and their
organisation from memory in general terms, uncertainty, reflected in Non

Uncertainty and sketching behaviour 477


Part by Part production, may reappear as the components and their relation
to each other are more fully detailed (hence the increase in Non Part by
Part production after Drawing 2 and beyond). As the task progresses and
the designers become increasingly committed to their solutions, there may
be phases when drawings are produced with a high degree of certainty
(e.g. Drawing 6).

5.5 Why does uncertainty engender cognitive shift?


All of the above leaves open the question as to why uncertainty should
yield Non Part by Part production, that is, what is the nature of the cogni-
tive shift. If the information necessary for visualisation is either not present
or difficult to recall then this may cause uncertainty. To resolve this lack,
the designers must dig deep into memory either to recall stored task-chair
information, or that associated with other previously seen objects. In other
words they must somehow fill in the holes in their understanding of chair.

Perhaps the simplest way of explaining why parts are drawn is to assume
a cognitive focusing mechanism allowing attention to be spotlighted on
areas of the mental representation involved in the task23. Under conditions
of certainty this model is scanned component by component. If the model
is incomplete, then it may be possible to focus attention down onto the
area of uncertainty to produce a temporary representation in which the
‘hole’ is somehow defined by those parts of the model that impinge on it.
However, it may be that this localised focusing limits access to resolved
areas of the model. As the area of uncertainty is filled it may become
available to visualisation. The act of focusing might be beneficial both in
searching for stored task-chair or other object information. Of course, this
is a speculation that needs to be tested empirically. Nevertheless, if we
accept that Non Part by Part production reflects cognitive uncertainty, it
is interesting to note that the designers invariably work from the unknown
to the known. They do not draw the resolved fragments of parts, leaving
the ‘holes’ to be filled in, but resolve the ‘hole’ and then complete the parts.

6 The role of drawing stimuli in uncertainty


Returning to the possible explanations for the switching between Part by
Part and Non Part by Part set out in the second section of the paper, while
favouring top-down cognitive factors it is likely that such change is influ-
enced by drawing stimuli. We focus below on two ways by which drawing
stimuli probably contribute to the task completion.

23 Kosslyn, S M Image and 6.1 Engendering uncertainty


Brain: The Resolution of the Ima- Drawing stimuli may contribute to uncertainty. As designers produce chairs
gery Debate MIT Press, Cam-
bridge, MA (1994) from memory they appear to be testing what they have reproduced (e.g.

478 Design Studies Vol 21 No 5 September 2000


by matching external depictions to memories). Mismatches may be experi-
enced without a clear sense initially of what is actually wrong, engendering
uncertainty to be resolved. This is like the situation where you recognise
a friend but also detect that he looks different—and you realise he has
shaved off his beard. Figure 5(f) is arguably an attempt to resolve uncer-
tainty arising from a mismatch between external representations and mem-
ory. Initially the designer draws the entire chair Part by Part, but with a
back-leg component not present in the actual chair. The designer may
experience a mismatch between the way the back legs of the depicted chair
and remembered chair connect to the seat. In Figure 5(f) the designer cor-
rects this error, but only partially as he still retains a member that connects
the seat to the back.

The persistence of this back-member error perhaps reflects the outcome of


another kind of test in which designers appear to reason about the func-
tionality of their depictions. When describing Figure 5(b) above, we prom-
ised to return to consider why the designer should have discarded what
was in fact a correct recollection for an incorrect invention. If we assume
that the participant was not deliberately designing the part, then it could
be that the designer reasoned that the chair, as recollected, would not func-
tion and therefore opted for a vaguer but more functional recollection.

Figure 6 provides clearer evidence for such reasoning processes and their
role in creating uncertainty. Figure 6 shows five depictions of the School
chair. In D1 all of the parts are externalised. However, the leg members
are wrong as they show front-back leg members. The designer then pro-
duces D2–D4, but seems to change his mind between D1 and D4 (D3
displaying Non Part by Part production) as the legs are now shown bending
toward the middle of the seat. Why might this change have occurred?
Considering D1, what is the function of the bolts? They appear to have
none as they are out of alignment with the legs. In D2 the relation between
bolts and legs is more salient than in D1. At this point the designer may
have realised that the legs would have to be positioned such as to bolt to
the seat, i.e. D4. Finally, in D5 the designer first draws the seat and the
legs, occluded by the seat. At this point the designer appears uncertain
about the legs and draws in the occluded parts, but this time showing them
correctly as a front and a back pair. Furthermore, although the front legs
(cf. Figure 1(b)) are shown incorrectly as bent into the horizontal plane,
this detail suggests that the change of mind reflects increased recollection
of the actual chair rather than simply an improved design. Thus, drawing
stimuli may engender uncertainty when compared to memory and when
used for reasoning about the functionality or logical coherence of depic-
tions.

Uncertainty and sketching behaviour 479


Figure 6 Reasoning with sketches

6.2 Facilitating the resolution of uncertainty


We have suggested that uncertainty arises when internal representations are
inadequate for the task in hand. Resolution of the problem would involve a
process in which the missing information is recovered or constructed from
prior knowledge. Resolution of uncertainty will depend on how success-
fully memory and synthesis processes are stimulated. Along the lines sug-
gested by Fish and Scrivener5, it may be that representations of external
stimuli can be fused with internally generated representations around the
corresponding area of internal uncertainty. By stimulating recognition and
recall processes, these externally derived but incomplete data might facili-
tate a stream of mental images. Whatever the mechanism, the assumption
is that drawing stimuli facilitate the resolution of uncertainty by providing
cues or clues that stimulate the processes essential to it (i.e. additional
information). Whilst these data cannot shed further light on this proposition
it is notable that many decades of research on human memory have pro-
vided evidence for partial or incorrect memory productions stimulating cor-
rect recall (e.g. as in the tip-of-the-tongue phenomenon) and for external
cues similarly promoting improved recall19.

7 Conclusion
Because idea sketching is partially determined by underlying cognitive
structures and processes12,13,15 we can learn much about those processes

480 Design Studies Vol 21 No 5 September 2000


by studying its structure. It has been argued that uncertainty is the primary
determinant of the Non Part by Part sketching production reported by
Kavakli et al.15. If this descriptive analysis is confirmed experimentally,
Non Part by Part production may prove a useful metric of uncertainty in
form making. While it does not appear to determine Non Part by Part
production, perception of drawing stimuli may contribute both to engen-
dering and resolving uncertainty and drawing stimuli probably facilitate
visual reasoning.

Drawing in design and recall may be different matters but their correspon-
dences are enough to suggest that the results may generalise. First, recalling
a previously seen object draws on episodic memory which is a feature of
design thinking6,24. Second, participants’ sketches are incomplete5, and like
idea sketches probably function to stimulate recall and inhibit premature
commitment5. However, here we are suggesting uncertainty is the cause
of incompleteness, i.e. its production is not a deliberate strategy exploiting
perception to stimulate invention. The same is quite likely to be the case in
design also. Third, like design, sketches seem to support reinterpretation5–8
but here revealing weaknesses in conception that cause backward reflection
on the current conception rather than highlighting new conceptions that
move the task forward. Again reinterpretation may also contribute in this
way to design. Finally, the recall drawing task becomes a design task,
i.e. requiring invention, when episodic memory is insufficient for concrete
specification. This is probably the general case for any visual recall task
requiring the reification of recalled entities. Here, it is argued that uncer-
tainty drives invention and perhaps uncertainty also drives invention in
design.
24 Visser, W ‘Use of episodic
knowledge an information in Acknowledgments
design problem solving’ in N
Cross, H Christians and K To Dr Manolya Kavakli of the Faculty of Architecture, Istanbul Technical
Dorst (eds) Analysing Design University, Turkey, who worked with Scrivener and Ball on the design
Activity John Wiley and Sons,
Chichester (1996) pp 271–289 and execution of the study on which this paper is based.

Uncertainty and sketching behaviour 481

Vous aimerez peut-être aussi