Sketching is thinking: why the cheapest medium carries the most consequential moves

The cheapest thing a designer makes is usually the most consequential: a thumbnail takes seconds, is thrown away without regret, and can decide which problem the project is really about. Bill Buxton (2007) states the distinction as two aims, getting the design right and getting the right design. Refinement serves the first; the second needs candidates drawn before any is developed, because the costliest decision is which idea to pursue at all.

The sketch, this unit argues, is not a poor copy of a finished design. It is a different instrument. Its speed and ambiguity do cognitive work that a polished rendering cannot. The claim has a firm theoretical base and a thin empirical one. Nearly every observational result below comes from a protocol study of one or two designers. This is the fourth of ten units; it takes the modelled claim and hands unit five the candidates it has drawn.

The sketch is not a record of a thought

Goel (1995) supplies the theory. He argues that the computational picture of mind demands precise, unambiguous thought. That picture therefore leaves out dense, ambiguous symbol systems such as sketching, whose cognitive role is not trivial. He studies these processes through design problem solving. He holds that different symbol systems facilitate different cognitive processes. The notation is part of the reasoning, not a neutral surface for it.

Goldschmidt (1991) gives the observational version. Her protocol study of an architect found that sketching did not represent images already held in mind. It produced visual displays that induced images of the entity being designed. A regular oscillation ran between arguments about figural properties and arguments about non-figural matters such as function. She called this the dialectics of sketching. The oscillation ends when the designer judges the drawing coherent.

A sequence reaches what one drawing cannot

Goldschmidt (1992) extends the mechanism to a run of drawings. In serial sketching the designer transforms partial images of an entity that does not yet exist. Each sketch provides feedback that informs the next.

Goldschmidt (2014) models the picture. The sketch is a rapid external representation. An experienced sketcher produces it at next to no cognitive cost. It tolerates incompletion, inaccuracy and lack of scale, and a designer can easily revise it. Working in tandem with internal imagery, it is a strategic skill worth mastering even in the digital age.

The sketch returns more than it was given

Suwa and Tversky (1997) asked what designers read off their own freehand drawings. Advanced design students and practising architects designed while drawing freehand, were interviewed afterwards about what they had been thinking, and short stretches of the recordings were analysed in detail. The transcripts show a reading that exceeds intention: the authors report that designers discovered spaces and things in configurations of lines they had laid down without meaning them, and read functional matters, such as how a room would be used, from the visual attributes of shapes and angles.

Suwa, Purcell and Gero (1998) built the instrument that makes it inspectable. Their scheme codes designers' actions from video and audio protocols into four levels: physical, perceptual, functional and conceptual. Applied to one practising architect, it showed that sketches serve three roles. They are external memory, a provider of visual cues for association, and a setting in which design thoughts are constructed on the fly. That is the thesis stated most sharply and demonstrated thinnest: one designer.

Purcell and Gero (1998) supply the mechanism. They review protocol studies beside cognitive psychology on imagery reinterpretation. They place the ambiguous, unstructured sketch early in the process, where a designer can read an ambiguous figure again differently and produce features they did not deliberately draw.

The medium affords nothing until the reader brings expertise

Menezes and Lawson (2006) put pressure on the medium-side account. They compared how novice and advanced design students describe conceptual sketches to a peer. They found differences between the groups in their use of formal and symbolic references. What a drawing affords depends on what the reader brings, and the reader these studies assume can see.

Cross (2004) finds the same asymmetry. He reviews expert performance in design. He reports that experienced designers are solution-focused rather than problem-focused, and that they move rapidly to early conjectures and frame the problem proactively. The consequential move is the designer's framing, and the sketch only carries its traces.

Three studies measure an outcome, and none isolates the medium

Three studies here measure something a designer or client could inspect, not only process.

Schutze, Sachse and Romer (2003) compared design problem solving done entirely mentally, partly supported by sketching, and completely supported. The quality of the solution concepts increased across the three conditions, with continuous movement between internal mental activity and external materialising activity. It is the strongest experimental support here for externalisation, though it isolates externalisation rather than the sketch's ambiguity.

Yang (2009) observed engineering designers generating concepts by brainstorming, morphology charts and sketching. She found statistically significant correlations between the quantity of brainstormed ideas and design outcome. Some experiments, but not all, found a correlation between morphological alternatives and outcome. Dimensioned drawings in the early-to-middle phases also correlated with outcome, which points to concrete sketching and timing.

Bilda and Demirkan (2003) compared the media directly. Six interior designers solved the same interior space-planning problem in an office setting and again in a computer laboratory. Their protocols were coded by this tradition's cognitive action scheme. Traditional media showed advantages over the digital in the perception of visual-spatial features, in the production of alternative solutions, and in the conception of the problem. Six designers carry the whole claim.

Fallman (2003) supplies the frame in which that matters for the field. He argues that HCI is a design-oriented research field. In it the designer works problem and solution out as a closely coupled pair. He also argues that design-oriented research produces knowledge by making artefacts. The sketch is where that coupling is visible.

Contested question: does the medium think, or does the designer?

One position holds that the external representation does the cognitive work. Goldschmidt (1991, 1992), Purcell and Gero (1998) and Suwa, Purcell and Gero (1998) treat the medium's properties as causal. Ambiguity and incompletion let the designer notice something and reframe. The sketch is a physical setting in which design thoughts are constructed on the fly. A smoother medium would remove part of the mechanism.

The other holds that the designer does the work and the medium only records it. Cross (2004) locates the distinctive move in the expert's problem framing. Menezes and Lawson (2006) show that novice and advanced students describe the same kind of sketch differently. Goldschmidt (2014) rests the efficiency of sketching on the sketcher's experience. A rendering in expert hands, or an expert working without a drawing, could carry the same moves.

The divergence is testable: hold the design task constant and vary, factorially, the ambiguity of the representation and the expertise of the designer, then measure both the features read off that were not deliberately drawn and the quality of the design that results. No study here does that, since Suwa, Purcell and Gero (1998) code that detail for one architect, Menezes and Lawson (2006) compare novices and advanced students but measure description rather than outcome, and Yang (2009) relates drawing volume to outcome without manipulating either factor.

A second disagreement sits underneath: does the medium matter at all? Bilda and Demirkan (2003) found traditional media ahead of digital with six designers and no outcome measure. Goldschmidt (2014) recommends sketching in the digital age without testing one medium against another. Settling it needs the same task, matched expertise, more than six participants, and an outcome rather than a coding count.

The two sides rest on evidence that is thin on both. The medium-side case is protocol studies of one or two designers, so it observes a mechanism at a scale that cannot generalise, and the designer-side case is a review of expert performance and a comparison of how novices and advanced students describe the same sketch, which measures description rather than outcome. Only Schutze, Sachse and Romer manipulate the medium, and they isolate externalisation rather than ambiguity. The strongest objection is that the unit cannot show the sketch causes the move rather than recording it: an expert who thinks without drawing may reach the same reframing. While the dispute is open, a designer treats the drawing as one input to a move they must still make, and keeps the ambiguity that invites it.

Where the argument is heading

Two directions are visible in named work. The first moves from whether designers sketch to when and how concretely: Goldschmidt (2014) argues that sketching remains a strategic skill in the digital age, Yang (2009) finds that dimensioned drawing in the early-to-middle phases tracks the outcome, and Bilda and Demirkan (2003) remain the only protocol comparison of the digital transition. The second is the sketch as evidence, where Fallman (2003) argues that making artefacts produces knowledge, turning a studio drawing from a communication device into something that can support a claim. What the area lacks is the study separating the medium's contribution from the designer's.

The sources this unit rests on

• Bilda, Z. and Demirkan, H. (2003). An insight on designers sketching activities in traditional versus digital media. Design Studies. doi:10.1016/S0142-694X(02)00032-7 Bilda and Demirkan coded the protocols of six interior designers solving a space-planning problem through a transition between traditional and digital media. Traditional media showed advantages over the digital, supporting the production of alternative solutions and a better conception of the design problem.
• Buxton, B. (2007). Sketching User Experiences: Getting the Design Right and the Right Design. Morgan Kaufmann. https://archive.org/details/sketchinguserexp0000buxt Buxton argues that sketching and prototyping are separate activities with separate jobs: getting the design right, and getting the right design. A sketch is quick, disposable and ambiguous on purpose, so its roughness invites change where a rendering invites approval.
• Cross, N. (2004). Expertise in design: an overview. Design Studies. doi:10.1016/j.destud.2004.06.002 Cross reviews expert performance in design and reports that successful design behaviour rests on problem scoping rather than exhaustive problem analysis, with experts solution-focused and proactive in framing. Generating a range of alternatives is recommended by theorists but is not normal expert practice.
• Fallman, D. (2003). Design-oriented human-computer interaction. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI). doi:10.1145/642611.642652 Fallman argues that HCI is a design-oriented research field, examined through the role of sketching, in which the designer works problem and solution out as a closely coupled pair. He separates design-oriented research, which produces knowledge by making artefacts, from research-oriented design.
• Goel, V. (1995). Sketches of Thought. MIT Press. doi:10.7551/mitpress/6270.001.0001 Goel argues that the computational conception of mind demands precise and unambiguous thought, so it leaves out dense and ambiguous symbol systems such as sketching whose cognitive role is non-trivial. He treats design problem solving as the domain in which these inarticulate processes can be studied.
• Goldschmidt, G. (1991). The dialectics of sketching. Creativity Research Journal. doi:10.1080/10400419109534381 Goldschmidt's protocol study of an architect found that sketching did not represent images held in mind but produced visual displays that induced images of the entity being designed. She identified a regular oscillation between figural and non-figural arguments, the dialectics of sketching, ending when the designer judged the drawing coherent.
• Goldschmidt, G. (1992). Serial sketching: visual problem solving in designing. Cybernetics and Systems. doi:10.1080/01969729208927457 Goldschmidt analysed serial sketching as small-step transformations of partial images of an entity that does not yet exist, each sketch providing feedback that informs the next. She treats the sequence as an exemplar of information processing in problem solving.
• Goldschmidt, G. (2014). Modelling the role of sketching in design idea generation. An Anthology of Theories and Models of Design, Springer. doi:10.1007/978-1-4471-6338-1_21 Goldschmidt models sketching in the early search phase as a rapid external representation produced, by an experienced sketcher, at next to no cognitive cost. It tolerates incompletion, inaccuracy and lack of scale, is easily revised, and works in tandem with internal imagery.
• Menezes, A. and Lawson, B. (2006). How designers perceive sketches. Design Studies. doi:10.1016/j.destud.2006.02.001 Menezes and Lawson examined how novice and advanced design students describe conceptual sketches to a peer, finding differences between the groups in their use of formal and symbolic references. They also asked whether sketches from the describer's own domain are described differently from those of another.
• Purcell, A. T. and Gero, J. S. (1998). Drawings and the design process. Design Studies. doi:10.1016/S0142-694X(98)00015-5 Purcell and Gero review protocol studies of drawing in design and other disciplines alongside cognitive psychology on working memory, imagery reinterpretation and mental synthesis. They place the ambiguous, unstructured sketch early in the process, when reinterpretation is most useful.
• Schutze, M., Sachse, P. and Romer, A. (2003). Support value of sketching in the design process. Research in Engineering Design. doi:10.1007/s00163-002-0028-7 Schutze, Sachse and Romer compared entirely mental design problem solving with partly and completely sketching-supported solving. The quality of the solution concepts increased across the three conditions, with designers moving between internal mental activity and external materialising activity.
• Suwa, M. and Tversky, B. (1997). What do architects and students perceive in their design sketches? A protocol analysis. Design Studies. doi:10.1016/S0142-694X(97)00008-2 Suwa and Tversky used a protocol analysis of retrospective reports, in which participants designed while drawing freehand and later reported what they had been thinking. They asked what information designers read off their own sketches and how they interact with them.
• Suwa, M., Purcell, T. and Gero, J. (1998). Macroscopic analysis of design processes based on a scheme for coding designers cognitive actions. Design Studies. doi:10.1016/S0142-694X(98)00016-7 Suwa, Purcell and Gero devised a scheme coding designers' actions from video and audio protocols into physical, perceptual, functional and conceptual levels, with the dependencies between them. Applied to one practising architect, it found that sketches also serve as a physical setting in which design thoughts are constructed on the fly.
• Yang, M. C. (2009). Observations on concept generation and sketching in engineering design. Research in Engineering Design. doi:10.1007/s00163-008-0055-0 Yang observed engineering designers generating concepts through brainstorming, morphology charts and sketching, finding statistically significant correlations between the quantity of brainstormed ideas and design outcome. Dimensioned drawings in the early-to-middle phases also correlated with outcome, suggesting that concrete sketching and timing matter.