Observations of Concept Generation and
Sketching in Design
MARIA C. YANG
Daniel J. Epstein Dept. of Industrial & Systems Engineering
USC Viterbi School of Engineering
University of Southern California
3715 McClintock Avenue, GER 240C
Los Angeles, CA 90089
Concept generation is a key activity in the engineering design process. Design
wisdom holds that increased concept generation in the initial phases of design
leads to better design outcomes. One common way to represent design cognition
is by sketches. This paper examines concept generation as evidenced through
sketches as well as with morphology charts and brainstorming, and its potential
correlations with design outcome. Statistically significant correlations were found
between the quantity of dimensioned drawings generated in the early-to-middle
phases of design and design outcome. Likewise, the number design alternatives
enumerated from direct concept generation methods was found to have a
statistically significant correlation with design outcome.
Sketching, Concept Generation, Design Process
Concept generation is a key activity in the engineering design process. One
common method for visualizing concepts is sketching. This paper examines
concept generation as evidenced through sketches, morphology charts, and
brainstorming, and its potential correlations with design outcome.
One of the questions we address is, “Does the quantity of ideas generated in the
design process lead to a better design outcome?” [Osborn 1963] describes a
method of concept generation known as brainstorming, and suggests that coming
up with a larger number of ideas at the beginning of a process will lead to a better
resulting idea at the end of the process. By broadening the initial pool of ideas,
“quantity yields quality.” This practice has been widely adopted in product design
and development in industry. [Kelley and Littman 2001] hold that generating
more ideas through brainstorming leads to more creative design solutions.
Consider a design team that comes up with fifty initial design concepts, three of
which are considered “good” solutions. Then imagine if that same team is able to
ideate twice as many concepts. The team will presumably have increased the odds
of having more than three “good” solutions. Another way to assess concept
quality is by the concept’s novelty. If a roomful of designers was asked to
individually come up with design solutions to a problem, it could be reasonable to
assume that the first fifteen or twenty ideas that each designer generated might be
very similar. This initial set of ideas could be described as the most obvious
solutions. However, if each of these designers were able to generate one hundred
design solutions, it is more likely that each individual’s additional solutions would
differ from one another and that these additional solutions would probably be less
Concept generation is analyzed in two ways, implicitly through sketches, and then
explicitly through enumerated design alternatives. This paper builds on
preliminary research presented by the author in [Yang 2003].
First, ideas are observed through sketches created by designers. Sketching is an
integral part of the engineering design process and provides one form of evidence
of design activity. This paper looks at the quantity of design sketching to infer
concept generation volume. It also tracks the timing of concept generation and
sketching. Designers can, and do, form ideas throughout the design cycle.
Concepts may occur under formal circumstances (for example, to meet a
milestone) or more spontaneously, as a classic “eureka” moment. However,
concurrent engineering holds that design decisions made in the early, conceptual
phases of design have greater impact on overall design than those made later on
[Winner et al. 1988]. This study observes whether concepts generated earlier or
later on in the design cycle have different effects on design outcome.
The role of sketching is also considered. Sketches in this study are divided into
two categories: dimensioned drawings, and non-dimensioned drawings. It is
believed that, in some cases, dimensioned drawings represent ideas that are
further along in the design cycle than non-dimensioned ones.
Second, this paper examines concept generation more directly through the
quantity of ideas created through brainstorming and in morphology charts.
Brainstorming is free-form, unstructured concept generation activity. Morphology
charts are hierarchies of alternative embodiments that can fulfill each of a device’s
functions [Zwicky 1969]. These distinct design embodiments are taken to
represent explicit concept generation.
This paper draws on several existing areas of design research, including analysis
of design activity and research in sketching. This study examines the output of
design activity in order to gain insight into design thinking. The particular format
of design activity output investigated is sketching and concepts.
Analyzing Design Activity
A substantial body of research has focused on a descriptive approach to
understanding design activity and cognition through observation. Protocol
analysis focuses on the verbal output of design, and involves the observation of
designers “thinking-aloud” during design activity (Baya and Leifer 1994; Blessing
1995; Bucciarelli 1994; Cross, et al. 1996; Minneman and Leifer 1993; Tang and
Leifer 1991). By observing verbalized thoughts, researchers may better
understand designers’ process and rationale for decision-making.
Design activity may also be captured in the form of text and sketches generated by
designers and analyzed post-facto (Dong and Agogino 1996; Wood, et al. 1998;
Yang and Cutkosky 1999). This approach may be combined with interviews of
the designers themselves to gather additional context. The goal of this method is
again to understand the designer’s thinking, in particular how they handle
Other work has focused not only analyzing design activity but on linking design
activity with design outcome. (Dong, et al. 2002) found correlations between
textual design information coherence of design teams and design outcome.
(Mabogunje and Leifer 1996) investigated the association between the occurrence
of noun phrases in design documentation and team grades. This paper adopts a
similar correlation approach for linking sketching and concept generation in
design with design performance.
The Role of Sketching in Design
This paper holds that sketching as a design activity is closely linked with concept
generation and this notion is supported by other research. (Ullman, et al. 1990)
describe the importance of sketching as a way for designers to illustrate their
thinking. In fact, (Nagai and Noguchi 2003) show that designers constantly
translate non-visual textual specifications for a design into ones that are visual in
nature. The link between designing and visually representing artifacts through
sketches and prototypes has been discussed in depth by (McKim 1980) and
(Schön and Wiggins 1992). Observations of industrial practice suggest that design
success is closely linked to realizing an idea through drawing and prototyping
(Schrage and Peters 1999).
(Suwa and Tversky 1997) suggest drawing is not simply a representation of
design thought, but a critical vehicle for thinking about a design. They conclude
that designers are able to understand the various aspects of a design only through
sketching them. The work of several other researchers supports this idea that the
act of sketching is the same as design thinking. (Goel 1995) posits that sketches
are inherent in the cognitive activities that facilitate concept generation. (McKim
1980) maintains that the ability to “think visually” is a necessary skill for
developing innovative solutions. In a series of experiments, (Verstijnen, et al.
1998) concluded that certain sketching activities are inextricably linked to
creativity. (Cross 1999) proposes that sketching is a “dialogue” with the designer
in which his or her internal thinking interacts with the external representation of a
sketch. (Purcell and Gero 1998) suggest that the reinterpretation, or even iteration,
of a sketch is evidence of the occurrence of new conceptual design knowledge.
Indeed, (Tovey, et al. 2003) refers to sketching as a “language for handling design
ideas.” (Tovey, et al. 2003) also describes the evolution of sketches that results
over time “interactive generation.” (Goldschmidt 1994) describes the way in
which sketches are used both to represent ideas and to spur the generation of
Sketching serves other purposes besides cognition within the design process. First,
sketches serve as a flexible idea repository. (Goel 1995) argues that sketching is
critical to capturing the ambiguity inherent in design activity. (Kavakli, et al.
1998) refer to the ambiguity in conceptual design sketches as “partially
envisioned entities.” That is, a sketch is a way of preserving design freedom.
Sketches are also a communication tool. (Verstijnen, et al. 1998) points out that
sketches may function as a presentation medium to be share design information
It is worthwhile to consider not only the role of sketching in design, but the
various categories of sketches. (Ferguson 1992) classifies sketches in terms of
their intended purpose: the thinking sketch that is used by the designer alone as
reflective medium, the prescriptive sketch that is meant to serve as a blueprint for
those involved in the later stages of design, such as a draftsperson, and the talking
sketch which results from team members interactively sketching with each other
in order to communicate.
Sketches may also be categorized by the design progression represented. (Goel
1995) classifies the evolution of sketches as lateral or vertical transformations.
Lateral transformations are incremental changes that build on a previous idea,
while vertical ones result in a more detailed, thought out version of an earlier
sketch. Lateral transformations may be thought of as increasing breadth, while
vertical transformations increase depth.