What do Prototypes Prototype?

Stephanie Houde and Charles Hill

Apple Computer, Inc.
Cupertino, CA, USA
[email protected], [email protected]

1. INTRODUCTION the kinds of prototypes to build. With a clear pur-

Prototypes are widely recognized to be a core means pose for each prototype, we can better use proto-
of exploring and expressing designs for interactive types to think and communicate about design.
computer artifacts. It is common practice to build
prototypes in order to represent different states of In the first section we describe some current diffi-
an evolving design, and to explore options. How- culties in communicating about prototypes: the
ever, since interactive systems are complex, it may complexity of interactive systems; issues of multi-
be difficult or impossible to create prototypes of a disciplinary teamwork; and the audiences of pro-
whole design in the formative stages of a project. totypes. Next, we introduce the model and illus-
Choosing the right kind of more focused prototype trate it with some initial examples of prototypes
to build is an art in itself, and communicating its from real projects. In the following section we
limited purposes to its various audiences is a criti- present several more examples to illustrate some
cal aspect of its use. further issues. We conclude the chapter with a sum-
mary of the main implications of the model for
The ways that we talk, and even think about pro- prototyping practice.
totypes, can get in the way of their effective use.
Current terminology for describing prototypes cen- 2. THE PROBLEM WITH PROTOTYPES
ters on attributes of prototypes themselves, such as Interactive computer systems are complex. Any
what tool was used to create them, and how re- artifact can have a rich variety of software, hard-
fined-looking or -behaving they are. Such terms ware, auditory, visual, and interactive features. For
can be distracting. Tools can be used in many dif- example, a personal digital assistant such as the
ferent ways, and detail is not a sure indicator of Apple Newton has an operating system, a hard case
completeness. with various ports, a graphical user interface and
audio feedback. Users experience the combined
We propose a change in the language used to talk effect of such interrelated features; and the task of
about prototypes, to focus more attention on fun- designing—and prototyping—the user experience
damental questions about the interactive system is therefore complex. Every aspect of the system
being designed: What role will the artifact play in must be designed (or inherited from a previous sys-
a user’s life? How should it look and feel? How tem), and many features need to be evaluated in
should it be implemented? The goal of this chapter combination with others.
is to establish a model that describes any prototype
in terms of the artifact being designed, rather than Prototypes provide the means for examining de-
the prototype’s incidental attributes. By focusing sign problems and evaluating solutions. Selecting
on the purpose of the prototype—that is, on what the focus of a prototype is the art of identifying the
it prototypes—we can make better decisions about most important open design questions. If the arti-
fact is to provide new functionality for users—and
This article is published, in a different format, as Houde, S., thus play a new role in their lives—the most im-
and Hill, C., What Do Prototypes Prototype?, in Handbook of portant questions may concern exactly what that
Human-Computer Interaction (2nd Ed.), M. Helander,
T. Landauer, and P. Prabhu (eds.): Elsevier Science B. V:
role should be and what features are needed to sup-
Amsterdam, 1997. port it. If the role is well understood, but the goal

1
of the artifact is to present its functionality in a which shows a scenario of something being used, a
novel way, then prototyping must focus on how prototype.
the artifact will look and feel. If the artifact’s func-
tionality is to be based on a new technique, ques- The organization supporting a design project may
tions of how to implement the design may be the have an overly narrow expectation of what a pro-
focus of prototyping efforts. totype is. Shrage (1996) has shown that organiza-
tions develop their own “prototyping cultures”
Once a prototype has been created, there are sev- which may cause them to consider only certain kinds
eral distinct audiences that designers discuss pro- of prototypes to be valid. In some organizations,
totypes with. They are: the intended users of the only prototypes which act as proof that an artifact
artifact being designed; their design teams; and the can be produced are respected. In others, only
supporting organizations that they work within highly detailed representations of look and feel are
(Erickson, 1995). Designers evaluate their options well understood.
with their own team by critiquing prototypes of
alternate design directions. They show prototypes Is a brick a prototype? The answer depends on
to users to get feedback on evolving designs. They how it is used. If it is used to represent the weight
show prototypes to their supporting organizations and scale of some future artifact, then it certainly
(such as project managers, business clients, or pro- is: it prototypes the weight and scale of the artifact.
fessors) to indicate progress and direction. This example shows that prototypes are not neces-
sarily self-explanatory. What is significant is not
It is difficult for designers to communicate clearly what media or tools were are used to create them,
about prototypes to such a broad audience. It is but how they are used by a designer to explore or
challenging to build prototypes which produce feed- demonstrate some aspect of the future artifact.
back from users on the most important design ques-
tions. Even communication among designers re- 2.2 Current terminology
quires effort due to differing perspectives in a multi- Current ways of talking about prototypes tend to
disciplinary design team. Limited understanding focus on attributes of the prototype itself, such as
of design practice on the part of supporting orga- which tool was used to create it (as in “C”, “Direc-
nizations makes it hard for designers to explain their tor™”, and “paper” prototypes); and on how fin-
prototypes to them. Finally, prototypes are not self- ished-looking or -behaving a prototype is (as in
explanatory: looks can be deceiving. Clarifying “high-fidelity” and “low-fidelity” prototypes). Such
what aspects of a prototype correspond to the even- characterizations can be misleading because the ca-
tual artifact—and what don’t—is a key part of suc- pabilities and possible uses of tools are often mis-
cessful prototyping. understood and the significance of the level of fin-
ish is often unclear, particularly to non-designers.
2.1 What is a prototype?
Designing interactive systems demands collabo- Tools can be used in many different ways. Some-
ration between designers of many different dis- times tools which have high-level scripting lan-
ciplines (Kim, 1990). For example, a project might guages (like HyperCard™), rather than full pro-
require the skills of a programmer, an interaction gramming languages (like C), are thought to be
designer, an industrial designer, and a project man- unsuitable for producing user-testable prototypes.
ager. Even the term “prototype” is likely to be However, Ehn and Kyng (1991) have shown that
ambiguous on such a team. Everyone has a even prototypes made of cardboard are very useful
different expectation of what a prototype is. for user testing. In the authors’ experience, no one
Industrial designers call a molded foam model a tool supports iterative design work in all of the im-
prototype. Interaction designers refer to a simula- portant areas of investigation. To design well, de-
tion of on-screen appearance and behavior as a pro- signers must be willing to use different tools for
totype. Programmers call a test program a proto- different prototyping tasks; and to team up with
type. A user studies expert may call a storyboard, other people with complementary skills.

2
Finished-looking (or -behaving) prototypes are of- Role
ten thought to indicate that the design they repre-
sent is near completion. Although this may some-
times be the case, a finished-looking prototype
might be made early in the design process (e.g., a
3D concept model for use in market research), and
a rough one might be made later on (e.g., to em-
phasize overall structure rather than visual details
in a user test). Two related terms are used in this Implementation
context: “resolution” and “fidelity”. We interpret
resolution to mean “amount of detail”, and fidel- Look and feel

ity to mean “closeness to the eventual design”. It is Figure 1. A model of what prototypes prototype.
important to recognize that the degree of visual and
behavioral refinement of a prototype does not nec-
essarily correspond to the solidity of the design, or inherently more important than any other.
to a particular stage in the process.

3. A MODEL OF WHAT PROTOTYPES PROTOTYPE Goal of the model

3.1 Definitions Given a design problem (of any scope or size), de-
Before proceeding, we define some important terms. signers can use the model to separate design issues
We define artifact as the interactive system being into three classes of questions which frequently de-
designed. An artifact may be a commercially re- mand different approaches to prototyping. Imple-
leased product or any end-result of a design activ- mentation usually requires a working system to be
ity such as a concept system developed for research built; look and feel requires the concrete user ex-
purposes. We define prototype as any representa- perience to be simulated or actually created; role
tion of a design idea, regardless of medium. This requires the context of the artifact’s use to be es-
includes a preexisting object when used to answer tablished. Being explicit about what design ques-
a design question. We define designer as anyone tions must be answered is therefore an essential aid
who creates a prototype in order to design, regard- to deciding what kind of prototype to build. The
less of job title. model helps visualize the focus of exploration.

3.2 The model Markers

The model shown in Figure 1 represents a three- A prototype may explore questions or design op-
dimensional space which corresponds to important tions in one, two or all three dimensions of the
aspects of the design of an interactive artifact. We model. In this chapter, several prototypes from real
define the dimensions of the model as role; look design projects are presented as examples. Their
and feel; and implementation. Each dimension cor- relationship to the model is represented by a marker
responds to a class of questions which are salient on the triangle. This is a simple way to put the
to the design of any interactive system. “Role” re- purpose of any prototype in context for the designer
fers to questions about the function that an artifact and their audiences. It gives a global sense of what
serves in a user’s life—the way in which it is useful the prototype is intended to explore; and equally
to them. “Look and feel” denotes questions about important, what it does not explore.
the concrete sensory experience of using an arti-
fact—what the user looks at, feels and hears while It may be noted that the triangle is a relative and
using it. “Implementation” refers to questions about subjective representation. A location toward one
the techniques and components through which an corner of the triangle implies simply that in the
artifact performs its function—the “nuts and bolts” designer’s own judgment, more attention is given
of how it actually works. The triangle is drawn to the class of questions represented by that corner
askew to emphasize that no one dimension is than to the other two.

3
 Role

3
Implementation 2

Look and feel

Figure 2. Relationship of three prototypes (Examples

1-3) to the model. Example 1. Role prototype for 3D space-planning
application [E1 Houde 1990].

3.3 Three prototypes of one system One of challenges of the project was to define an
The model is best explained further through an ex- easy-to-use direct manipulation user interface for
ample from a real project. The three prototypes moving 3D objects with an ordinary 2D mouse cur-
shown in Examples 1-3 were created during the sor. User testing with a foam-core model showed
early stages of development of a 3D space-planning that the most important manipulations of a space-
application (Houde, 1992). planning task were sliding, lifting, and turning fur-
niture objects. Example 2 shows a picture of a pro-
The goal of the project was to design an example totype which was made to test a user interface fea-
of a 3D application which would be accessible to a turing this constrained set of manipulations. Click-
broad range of nontechnical users. As such it was ing once on the chair object caused its bounding
designed to work on a personal computer with an box to appear. This “handle box” offered hand-
ordinary mouse. Many prototypes were created shaped controls for lifting and turning the box and
by different members of the multi-disciplinary de- chair (as if the chair was frozen inside the box).
sign team during the project. Clicking and dragging anywhere on the box allowed
the unit to slide on a 3D floor. The prototype was
The prototype shown in Example 1 was built to built using Macromedia Director (a high level ani-
show how a user might select furniture from an on- mation and scripting tool.) It was made to work
line catalog and try it out in an approximation of only with the chair data shown: a set of images pre-
their own room. It is an interactive slide show which drawn for many angles of rotation.
the designer operated by clicking on key areas of
the rough user interface. The idea that virtual space-
planning would be a helpful task for nontechnical
users came from user studies. The purpose of the
prototype was to quickly convey the proposed role
of the artifact to the design team and members of
the supporting organization.

Since the purpose of the prototype was primarily

to explore and visualize an example of the role of
the future artifact, its marker appears very near the
role corner of the model in Figure 2. It is placed a
little toward the look and feel corner because it also
explored user interface elements in a very initial Example 2. Look-and-feel prototype for 3D space-
form. planning application [E2 Houde 1990].

4
 the left side of the screen were made by the pro-
grammer to give himself controls for demonstrat-
ing the system: they were not made to explore the
look and feel of the future artifact. Thus the pri-
mary purpose of the prototype was to explore how
the artifact might be implemented. The marker for
this example is placed near the implementation cor-
ner (Figure 2).

One might assume that the role prototype (Example

1) was developed first, then the look and feel pro-
Example 3. Implementation prototype for 3D space- totype (Example 2), and finally the implementation
planning application [E3 Chen 1990]. prototype (Example 3): that is, in order of increas-
ing detail and production difficulty. In fact, these
three prototypes were developed almost in paral-
The purpose of Example 2 was to get feedback from lel. They were built by different design team mem-
users as quickly as possible as to whether the look bers during the early stages of the project. No single
and feel of the handle box user interface was prom- prototype could have represented the design of the
ising. Users of the prototype were given tasks which future artifact at that time. The evolving design
encouraged them to move the chair around a vir- was too fuzzy—existing mainly as a shared con-
tual room. Some exploration of role was supported cept in the minds of the designers. There were also
by the fact that the object manipulated was a chair, too many open and interdependent questions in
and space-planning tasks were given during the test. every design dimension: role, look and feel, imple-
Although the prototype was interactive, the pro- mentation.
gramming that made it so did not seriously explore
how a final artifact with this interface might be imple- Making separate prototypes enabled specific design
mented. It was only done in service of the look and questions to be addressed with as much clarity as
feel test. Since the designer primarily explored the possible. The solutions found became inputs to an
look and feel of the user interface, this prototype’s integrated design. Answers to the rendering capa-
marker is placed very near the look and feel corner bility questions addressed by Example 3 informed
of the model in Figure 2. the design of the role that the artifact could play
(guiding how many furniture objects of what com-
A technical challenge of the project was figuring plexity could be shown). It also provided guiding
out how to render 3D graphics quickly enough on constraints for the direct manipulation user inter-
equipment that end-users might have. At the time, face (determining how much detail the handle forms
it was not clear how much real time 3D interaction could have). Similarly, issues of role addressed by
could be achieved on the Apple Macintosh™ IIfx Example 1 informed the implementation problem
computer—the fastest Macintosh then available. by constraining it: only a constrained set of ma-
Example 3 shows a prototype which was built pri- nipulations was needed for a space-planning appli-
marily to explore rendering capability and perfor- cation. It also simplified the direct manipulation
mance. This was a working prototype in which user interface by limiting the necessary actions and
multiple 3D objects could be manipulated as in therefore controls which needed to be provided.
Example 2, and the view of the room could be
changed to any perspective. Example 3 was made It was more efficient to wait on the results of inde-
in a programming environment that best supported pendent investigations in the key areas of role, look
the display of true 3D perspectives during manipu- and feel and implementation than to try to build a
lation. It was used by the design team to determine monolithic prototype that integrated all features
what complexity of 3D scenes was reasonable to from the start. After sufficient investigation in sepa-
design for. The user interface elements shown on rate prototypes, the prototype in Example 3 began

5
 Role ity that a user might benefit from, with little atten-
tion to how the artifact would look and feel, or
Integration how it could be made to actually work. Designers
find such prototypes useful to show their design
teams what the target role of the artifact might be;
to communicate that role to their supporting orga-
nization; and to evaluate the role in user studies.

Implementation A portable notebook computer

The paper storyboard shown in Example 4 was an
Look and feel early prototype of a portable notebook computer
Figure 3. Four principal categories of prototypes on for students which would accept both pen and fin-
the model. ger input. The scenario shows a student making
notes, annotating a paper, and marking pages for
to evolve into an integrated prototype which could later review in a computer notebook. The designer
be described by a position at the center of our presented the storyboard to her design team to fo-
model. A version of the user interface developed cus discussion on the issues of what functionality
in Example 2 was implemented in the prototype in the notebook should provide and how it might be
Example 3. Results of other prototypes were also controlled through pen and finger interaction. In
integrated. This enabled a more complete user test terms of the model, this prototype primarily ex-
of features and user interface to take place. plored the role of the notebook by presenting a
rough task scenario for it. A secondary consider-
This set of three prototypes from the same project ation was a rough approximation of the user inter-
shows how a design problem can be simultaneously face. Its marker, shown in Figure 4, is therefore
approached from multiple points of view. Design positioned near the role corner of the model and a
questions of role, look and feel, and implementa- little toward look and feel.
tion were explored concurrently by the team with
the three separate prototypes. The purpose of the Storyboards like this one are considered to be ef-
model is to make it easier to develop and subse- fective design tools by many designers because they
quently communicate about this kind prototyping help focus design discussion on the role of an arti-
strategy. fact very early on. However, giving them status as
prototypes is not common because the medium is
4. FURTHER EXAMPLES paper and thus seems very far from the medium of
In this section we present twelve more examples of
prototypes taken from real projects, and discuss
them in terms of the model. Examples are divided
into four categories which correspond to the four
main regions of the model, as indicated in Figure
3. The first three categories correspond to proto-
types with a strong bias toward one of the three
corners: role, look and feel, and implementation
prototypes, respectively. Integration prototypes
occupy the middle of the model: they explore a
balance of questions in all three dimensions.

4.1 Role prototypes

Role prototypes are those which are built prima-
rily to investigate questions of what an artifact Example 4. Storyboard for a portable notebook
could do for a user. They describe the functional- computer [E4 Vertelney 1990].

6
 Role clicking on the voice tool brought up a picture of
6 some sound tools; and so on. To demonstrate the
prototype, the designer sat in front of a computer
7
4 and play-acted the role of a user opening her mail,
5 replying to it, and so forth. The prototype was used
in design team discussions and also demonstrated
to project managers to explain the current design
direction. According to the model, this prototype
Implementation primarily explored the role that certain features of
the operating system could play in a user’s daily
Look and feel tasks. It was also used to outline very roughly how
its features would be portrayed and how a user
Figure 4. Relationship of role prototypes (Examples
4-7) to the model. would interact with it. As in the previous example,
the system’s implementation was not explored. Its
marker is shown in Figure 4.
an interactive computer system. We consider this
storyboard to be a prototype because it makes a To make the prototype, user interface elements were
concrete representation of a design idea and serves hand-drawn and scanned in. Transitions between
the purpose of asking and answering design ques- steps in the scenario were made interactive in Mac-
tions. Of course, if the designer needed to evaluate romedia Director. This kind of portrayal of on-
a user’s reaction to seeing the notebook or to using screen interface elements as rough and hand-drawn
the pen-and-finger interaction, it would be neces- was used in order to focus design discussion on the
sary to build a prototype which supported direct overall features of a design rather than on specific
interaction. However, it might be wasteful to do so details of look and feel or implementation (Wong,
before considering design options in the faster, 1992). Ironically, while the design team understood
lighter-weight medium of pencil and paper. the meaning of the hand-drawn graphics, other
members of the organization became enamored with
An operating system user interface the sketchy style to the extent that they considered
Example 5 shows a screen view of a prototype that using it in the final artifact. This result was en-
was used to explore the design of a new operating tirely at odds with the original reasons for making
system. The prototype was an interactive story: it a rough-looking prototype. This example shows
could only be executed through a single, ordered, how the effectiveness of some kinds of prototypes
sequence of interactions. Clicking with a cursor may be limited to a specific kind of audience.
on the mailbox picture opened a mail window; then
The Knowledge Navigator
Example 6 shows a scene from Apple Computer’s
Knowledge Navigator™ video. The video tape tells
a day-in-the-life story of a professor using a futur-
istic notebook computer (Dubberly and Mitch,
1987). An intelligent agent named “Phil” acts as
his virtual personal assistant, finding information
related to a lecture, reminding him of his mother’s
birthday, and connecting him with other professors
via video-link. The professor interacts with Phil by
talking, and Phil apparently recognizes everything
said as well as a human assistant would.

Example 5. Interactive story for an operating system Based on the model, the Knowledge Navigator is
interface [E5 Vertelney and Wong 1990]. identified primarily as a prototype which describes

7
 normally be expected to understand such approxi-
mate representations.

The Integrated Communicator

Example 7 shows an appearance model of an Inte-
grated Communicator created for customer research
into alternate presentations of new technology (ID
Magazine 1995). It was one of three presentations
of possible mechanical configurations and interac-
tion designs, each built to the same high finish and
accompanied by a video describing on-screen inter-
actions. In the study, the value of each presenta-
Example 6. Knowledge Navigator™ vision video for a tion was evaluated relative to the others, as per-
future notebook computer [E6 Dubberly and Mitch ’87].
ceived by study subjects during one-on-one inter-
views. The prototype was used to help subjects
the role that the notebook would play in such a imagine such a product in the store and in their
user’s life. The story is told in great detail, and it is homes or offices, and thus to evaluate whether they
clear that many decisions were made about what would purchase such a product, how much they
to emphasize in the role. The video also shows spe- would expect it to cost, what features they would
cific details of appearance, interaction, and perfor- expect, etc.
mance. However, they were not intended by the
designers to be prototypes of look and feel. They The prototype primarily addresses the role of the
were merely placeholders for the actual design work product, by presenting carefully designed cues which
which would be necessary to make the product re- imply a telephone-like role and look-and-feel. Fig-
ally work. Thus its marker goes directly on the ure 4 shows its marker near the role corner of the
role corner (Figure 4). model. As with the Knowledge Navigator, the very
high-resolution look and feel was a means of mak-
Thanks to the video’s special effects, the scenario ing the design as concrete as possible to a broad
of the professor interacting with the notebook and audience. In this case however it also enabled a
his assistant looks like a demonstration of a real basic interaction design strategy to be worked out
product. Why did Apple make a highly produced and demonstrated. The prototype did not address
prototype when the previous examples show that a implementation.
rapid paper storyboard or a sketchy interactive pro-
totype were sufficient for designing a role and tell-
ing a usage story? The answer lies in the kind of
audience. The tape was shown publicly and to
Apple employees as a vision of the future of com-
puting. Thus the audience of the Knowledge Navi-
gator was very broad—including almost anyone in
the world. Each of the two previous role design
prototypes was shown to an audience which was
well informed about the design project. A rough
hand-drawn prototype would not have made the
idea seem real to the broad audience the video ad-
dressed: high resolution was necessary to help
people concretely visualize the design. Again, while
team members learn to interpret abstract kinds of Example 7. Appearance model for the Integrated
prototypes accurately, less expert audiences cannot Communicator [E7 Udagawa 1995].

8
The key feature of this kind of prototype is that it
is a concrete and direct representation, as visually
finished as actual consumer products. These at-
tributes encourage an uncoached person to directly
relate the design to their own environment, and to
the products they own or see in stores. High qual-
ity appearance models are costly to build. There
are two common reasons for investing in one: to
get a visceral response by making the design seem
“real” to any audience (design team, organization,
and potential users); and to verify the intended look
and feel of the artifact before committing to pro-
duction tooling. An interesting side-effect of this Example 8. Animation of the look and feel of a fashion
prototype was that its directness made it a power- design workspace [E8 Hill 1992].
ful prop for promoting the project within the orga-
nization.
This is an example of a look and feel prototype.
4.2 Look and Feel prototypes The virtue of the prototype was that it enabled a
Look and feel prototypes are built primarily to ex- novel user interface design to be developed with-
plore and demonstrate options for the concrete ex- out having first to implement complex underlying
perience of an artifact. They simulate what it would technologies. While the role was inherited from
be like to look at and interact with, without neces- existing fashion design practice, the prototype also
sarily investigating the role it would play in the user’s demonstrated new options offered by the new com-
life or how it would be made to work. Designers puter-based approach. Thus, Figure 5 shows its
make such prototypes to visualize different look and marker in the look and feel area of the model.
feel possibilities for themselves and their design
teams. They ask users to interact with them to see One issue with prototypes like this one is that inex-
how the look and feel could be improved. They perienced audiences tend to believe them to be more
also use them to give members of their supporting functional than they are just by virtue of being
organization a concrete sense of what the future shown on a computer screen. When this prototype
artifact will be like. was shown, the designers found they needed to take
great care to explain that the design was not imple-
A fashion design workspace mented.
The prototype shown in Example 8 was developed
to support research into collaboration tools for fash- Role
ion designers (Hill et al., 1993; Scaife et al, 1994).
A twenty-minute animation, it presented the con-
cept design for a system for monitoring garment
design work. It illustrated in considerable detail
the translation of a proven paper-based procedure
into a computer-based system with a visually rich, 8
direct manipulation, user interface. The prototype’s 9
main purposes were to confirm to the design team Implementation
that an engaging and effective look and feel could 10
be designed for this application, and to convince Look and feel
managers of the possibilities of the project. It was Figure 5. Relationship of the look and feel prototypes
presented to users purely for informal discussion. (Examples 8-10) to the model.

9
A learning toy
The “GloBall” project was a concept for a children’s
toy: a ball that would interact with children who
played with it. Two prototypes from the project
are shown, disassembled, in Example 9. The de-
sign team wanted the ball to speak back to kids
when they spoke to it, and to roll towards or away
from them in reaction to their movements. The
two prototypes were built to simulate these func-
tions separately. The ball on the left had a walkie-
talkie which was concealed in use. A hidden op- Example 10. Pizza-box prototype of an architect’s
computer [E10 Apple Design Project, 1992].
erator spoke into a linked walkie-talkie to simulate
the ball’s speech while a young child played with it.
Similarly, the ball on the right had a radio-controlled An architect’s computer
car which was concealed in use. A hidden operator This example concerned the design of a portable
remotely controlled the car, thus causing the ball to computer for architects who need to gather a lot of
roll around in response to the child’s actions. information during visits to building sites. One of
the first questions the designers explored was what
As indicated by the marker in Figure 5, both proto- form would be appropriate for their users. With-
types were used to explore the toy’s look and feel out much ado they weighted the pizza box shown
from a child’s viewpoint, and to a lesser extent to in Example 10 to the expected weight of the com-
evaluate the role that the toy would play. Neither puter, and gave it to an architect to carry on a site
seriously addressed implementation. The design- visit. They watched how he carried the box, what
ers of these very efficient prototypes wanted to know else he carried with him, and what tasks he needed
how a child would respond to a toy that appeared to do during the visit. They saw that the rectilinear
to speak and move of its own free will. They man- form and weight were too awkward, given the other
aged to convincingly simulate novel and difficult- materials he carried with him, and this simple in-
to-implement technologies such as speech and au- sight led them to consider of a softer form. As
tomotion, for minimal cost and using readily avail- shown by its marker, this is an example of a rough
able components. By using a “man behind the cur- look and feel prototype (Figure 5). Role was also
tain” (or “Wizard of Oz”) technique, the designers explored in a minor way by seeing the context that
were able to present the prototypes directly to chil- the artifact would be used in.
dren and to directly evaluate their effect.
The pizza box was a very efficient prototype. Spend-
ing virtually no time building it or considering op-
tions, the students got useful feedback on a basic
design question—what physical form would be best
for the user. From what they learned in their simple
field test, they knew immediately that they should
try to think beyond standard rectilinear notebook
computer forms. They began to consider many dif-
ferent options including designing the computer to
feel more like a soft shoulder bag.

4.3 Implementation Prototypes

Some prototypes are built primarily to answer tech-
nical questions about how a future artifact might
Example 9. Look and feel simulation prototypes for a actually be made to work. They are used to dis-
child’s toy [E9 Bellman et al, 1993]. cover methods by which adequate specifications for

10
the final artifact can be achieved—without having Role
to define its look and feel or the role it will play for
a user. (Some specifications may be unstated, and
may include externally imposed constraints, such
as the need to reuse existing components or pro-
duction machinery.) Designers make implementa-
tion prototypes as experiments for themselves and 11
the design team, to demonstrate to their organiza- 12
tion the technical feasibility of the artifact, and to Implementation
get feedback from users on performance issues.
Look and feel

A digital movie editor Figure 6. Relationship of implementation prototypes

Some years ago it was not clear how much (Examples 11 and 12 ) to the model.
interactivity could be added to digital movies play-
ing on personal computers. Example 11 shows a
picture of a prototype that was built to investigate members of the movie editor design team. When
solutions to this technical challenge. It was an ap- the audience understood that an implementation
plication, written in the C programming language design was being demonstrated, discussion was fo-
to run on an Apple Macintosh computer. It offered cused productively. At other times it became fo-
a variety of movie data-processing functionality cused on problems with the user interface, such as
such as controlling various attributes of movie play. the multiple cascading menus, which were hard to
The main goal of the prototype was to allow mark- control and visually confusing. In these cases, dis-
ing of points in a movie to which scripts (which cussion was less productive: the incidental user in-
added interactivity) would be attached. As indi- terface got in the way of the intentional implemen-
cated by the marker in Figure 6, this was primarily tation.
a carefully planned implementation prototype.
Many options were evaluated about the best way The project leader shared some reflections after this
to implement its functions. The role that the func- somewhat frustrating experience. He said that part
tions would play was less well defined. The visible of his goal in pursuing a working prototype alone
look and feel of the prototype was largely inciden- was to move the project through an organization
tal: it was created by the designer almost purely to that respected this kind of prototype more than
demonstrate the available functionality, and was not “smoke and mirrors” prototypes—ones which only
intended to be used by others. simulate functionality. He added that one problem
might have been that the user interface was neither
This prototype received varying responses when good enough nor bad enough to avoid misunder-
demonstrated to a group of designers who were not standings. The edit list, which allowed points to be
marked in movies, was a viable look and feel de-
sign; while the cascading menus were not. For the
audience that the prototype was shown to, it might
have been more effective to stress the fact that look
and feel were not the focus of the prototype; and
perhaps, time permitting, to have complemented this
prototype with a separate look and feel prototype
that explained their intentions in that dimension.

A fluid dynamics simulation system

Example 12 shows a small part of the C++ pro-
Example 11. Working prototype of a digital movie gram listing for a system for simulating gas flows
editor [E11 Degen, 1994]. and combustion in car engines, part of an engineer-

11
 terfaces may never be properly redesigned before
IntList& IntList::operator=(const IntList& oldList) the final system is released. For these reasons it is
{ often desirable to treat even implementation pro-
register long n = oldList.size; totypes as disposable, and to migrate successful
if (n != size) setSize(n); implementation designs to a new integrated proto-
register int* newPtr = &values[n]; type as the project progresses.
register int* oldPtr = &oldList.values[n];
while (n--) *--newPtr = *--oldPtr; 4.4 Integration prototypes
return *this; Integration prototypes are built to represent the
} complete user experience of an artifact. Such pro-
totypes bring together the artifact’s intended design
Example 12. C++ program sample from a fluid in terms of role, look and feel, and implementa-
dynamics simulation system [E12 Hill, 1993].
tion. Integrated prototypes help designers to bal-
ance and resolve constraints arising in different de-
ing research project (Hill, 1993). One goal of this sign dimensions; to verify that the design is com-
prototype was to demonstrate the feasibility of ob- plete and coherent; and to find synergy in the de-
ject-oriented programming using the C++ language sign of the integration itself. In some cases the in-
in place of procedural programs written in the older tegration design may become the unique innova-
FORTRAN language. Object-oriented program- tion or feature of the final artifact. Since the user’s
ming can in theory lead to increased software re- experience of an artifact ultimately combines all
use, better reliability and easier maintenance. Since three dimensions of the model, integration proto-
an engine simulation may take a week to run on types are most able to accurately simulate the final
the fastest available computers and is extremely artifact. Since they may need to be as complex as
memory-intensive, it was important to show that the final artifact, they are the most difficult and
the new approach did not incur excessive perfor- time consuming kinds of prototypes to build. De-
mance or memory overheads. The program listing signers make integration prototypes to understand
shown was the implementation of the operation to the design as a whole, to show their organizations
copy one list of numbers to another. When tested, a close approximation to the final artifact, and to
it was shown to be faster than the existing get feedback from users about the overall design.
FORTRAN implementation. The prototype was
built primarily for the design team’s own use, and The Sound Browser
eventually used to create a deployable system. The The “SoundBrowser” prototype shown in Example
marker in Figure 6 indicates that this prototype pri- 13 was built as part of a larger project which inves-
marily explored implementation. tigated uses of audio for personal computer users
(Degen et al, 1992). The prototype was built in C
Other kinds of implementation prototypes include
demonstrations of new algorithms (e.g., a graphi- Role
cal rendering technique or a new search technol-
ogy), and trial conversions of existing programs to
run in new environments (e.g., converting a program
written in the C language to the Java language).
14
Implementation prototypes can be hard to build, 13 15
and since they actually work, it is common for them
to find their way directly into the final system. Two Implementation
problems arise from this dynamic: firstly, programs
developed mainly to demonstrate feasibility may Look and feel
turn out in the long term to be difficult to maintain Figure 7. Relationship of integration prototypes
and develop; and secondly, their temporary user in- (Examples 13-15) to the model.

12
to run on a Macintosh. It allowed a user to browse
digital audio data recorded on a special personal
tape recorder equipped with buttons for marking
points in the audio. The picture shows the
SoundBrowser’s visual representation of the audio
data, showing the markers below the sound dis-
play. A variety of functions were provided for re-
viewing sound, such as high-speed playback and
playback of marked segments of audio.
Example 14. Integration prototype of the “pile”
metaphor for information retrieval [E14 Rose, 1993].
This prototype earns a position right in the center
of the model, as shown in Figure 7. All three di-
mensions of the model were explored and repre- into the integrated prototype. This example shows
sented in the prototype. The role of the artifact the value of using different tools for different kinds
was well thought-out, being driven initially by ob- of design exploration, and how even at the end of a
servations of what users currently do to mark and project simple, low-fidelity prototypes might be
play back audio, and then by iteratively designed built to solve specific problems.
scenarios of how it might be done more efficiently
if electronic marking and viewing functions were The Pile Metaphor
offered. The look and feel of the prototype went The prototype shown in Example 14 was made as
through many visual design iterations. The imple- part of the development of the “pile” metaphor—a
mentation was redesigned several times to meet the user interface element for casual organization of
performance needs of the desired high-speed play- information (Mander et al, 1992, Rose et al, 1993).
back function. It represented the integration of designs developed
in several other prototypes which independently
When the SoundBrowser was near completion it explored the look and feel of piles, “content-aware”
was prepared for a user test. One of the features information retrieval, and the role that piles could
which the design team intended to evaluate was the play as a part of an operating system. In the pile
visual representation of the sound in the main win- metaphor, each electronic document was repre-
dow. They wanted to show users several alterna- sented by a small icon or “proxy”, several of which
tives to understand their preferences. The program- were stacked to form a pile. The contents of the
mer who built the SoundBrowser had developed pile could be quickly reviewed by moving the ar-
most of the alternatives. In order to refine these row cursor over it. While the cursor was over a
and explore others, two other team members cop- particular document, the “viewing cone” to the right
ied screen-shots from the tool into a pixel-painting displayed a short text summary of the document.
application, where they experimented with modifi-
cations. This was a quick way to try out different This prototype was shown to designers, project
visual options, in temporary isolation from other managers, and software developers as a proof of
aspects of the artifact. It was far easier to do this in concept of the novel technology. The implementa-
a visual design tool than by programming in C. tion design in this prototype might have been
When finished, the new options were programmed achieved with virtually no user interface: just text
input and output. However, since the prototype
was to be shown to a broad audience, an integrated
style of prototype was chosen, both to communi-
cate the implementation point and to verify that
the piles representation was practically feasible. It
helped greatly that the artifact’s role and look and
Example 13. Integrated prototype of a sound browser feel could be directly inherited from previous pro-
[E13 Degen 1993]. totypes. Figure 7 shows its marker on the model.

13
A garment history browser Role
The prototype in Example 15 was a working sys- 6
tem which enabled users to enter and retrieve snip- 1
7
pets of information about garment designs via a 4
visually rich user interface (Hill et al, 1993; Scaife 5
et al, 1994). The picture shows the query tool which
14
was designed to engage fashion designers and pro- 11 13 15 8
vide memorable visual cues. The prototype was 3 9
12
designed for testing in three corporations with a 2
Implementation
limited set of users’ actual data, and presented to 10
users in interviews. It was briefly demonstrated, Look and feel
then users were asked to try queries and enter re- 1. 3D space-planning (role)
marks about design issues they were currently aware 2. 3D space-planning (look and feel)
of. 3. 3D space-planning (implementation)
4. Storyboard for portable notebook computer
5. Interactive story, operating system user interface
This prototype was the end-result of a progression 6. Vision video, notebook computer
from an initial focus on role (represented by verbal 7. Appearance model, integrated communicator
usage scenarios), followed by rough look and feel 8. Animation, fashion design workspace
9. Look and feel simulation, child’s toy
prototypes and an initial implementation. Along 10. Pizza-box, architect’s computer
the way various ideas were explored, refined or re- 11. Working prototype, digital movie editor
jected. The working tool, built in Allegiant 12. C++ program listing, fluid dynamics simulation
13. Integrated prototype, sound browser
SuperCard™, required two months’ intensive work
14. Integrated prototype, pile metaphor
by two designers. In retrospect the designers had 15. Integrated prototype, garment history browser
mixed feelings about it. It was highly motivating
to users to be able to manipulate real user data Figure 8. Relationship of all examples to the model.
through a novel user interface, and much was
learned about the design. However, the designers
also felt that they had to invest a large amount of 5. SUMMARY
time in making the prototype, yet had only been In this chapter, we have proposed a change in the
able to support a very narrow role compared to the language used by designers to think and talk about
breadth shown in the animation shown in Example prototypes of interactive artifacts. Much current
8. Many broader design questions remained unan- terminology centers on attributes of prototypes
swered. themselves: the tools used to create them, or how
refined-looking or -behaving they are. Yet tools
can be used in many different ways, and resolution
can be misleading. We have proposed a shift in
attention to focus on questions about the design of
the artifact itself: What role will it play in a users
life? How should it look and feel? How should it
be implemented? The model that we have intro-
duced can be used by designers to divide any de-
sign problem into these three classes of questions,
each of which may benefit from a different approach
to prototyping.
We have described a variety of prototypes from real
projects, and have shown how the model can be
used to communicate about their purposes. Sev-
Example 15. Integrated prototype of a garment eral practical suggestions for designers have been
history browser [E15 Hill and Kamlish 1992]. raised by the examples:

14
• Define “prototype” broadly. Efficient prototypes use prototypes to think and communicate about
produce answers to their designers’ most important design.
questions in the least amount of time. Sometimes
very simple representations make highly effective 6. ACKNOWLEDGMENTS
prototypes: e.g., the pizza-box prototype of an Special thanks are due to Thomas Erickson for guid-
architect’s computer [Example 10] and the story- ance with this chapter, and to our many colleagues
board notebook [Example 1]. We define a proto- whose prototypes we have cited, for their comments
type as any representation of a design idea—regard- on early drafts. We would also like to acknowl-
less of medium; and designers as the people who edge S. Joy Mountford whose leadership of the
create them—regardless of their job titles. Human Interface Group at Apple created an atmo-
sphere in which creative prototyping could flour-
• Build multiple prototypes. Since interactive arti- ish. Finally, thanks to James Spohrer, Lori Leahy,
facts can be very complex, it may be impossible to Dan Russell, and Donald Norman at Apple Re-
create an integrated prototype in the formative search Labs for supporting us in writing this chap-
stages of a project, as in the 3D space-planning ex- ter.
ample [Examples 1, 2, and 3]. Choosing the right
focused prototypes to build is an art in itself. Be 6.1 Prototype Credits
prepared to throw some prototypes away, and to We credit here the principal designer and design
use different tools for different kinds of prototypes. team of each example prototype shown.

• Know your audience. The necessary resolution [E1] Stephanie Houde, [E2] Stephanie Houde, [E3]
and fidelity of a prototype may depend most on the Michael Chen. © Apple Computer, Inc., 1990. Project
team: Penny Bauersfeld, Michael Chen, Lewis Knapp
nature of its audience. A rough role prototype such
(project leader), Laurie Vertelney and Stephanie
as the interactive storyboard [Example 4] may work Houde.
well for a design team but not for members of the
supporting organization. Broader audiences may [E4] Laurie Vertelney. © Apple Computer Inc., 1990.
require higher-resolution representations. Some or- Project team: Michael Chen, Thomas Erickson, Frank
ganizations expect to see certain kinds of proto- Leahy, Laurie Vertelney (project leader).
types: implementation designs are often expected
[E5] Laurie Vertelney and Yin Yin Wong. © Apple
in engineering departments, while look-and-feel and Computer Inc., 1990. Project team: Richard Mander,
role prototypes may rule in a visual design envi- Gitta Salomon (project leader), Ian Small, Laurie
ronment. Vertelney, Yin Yin Wong.

• Know your prototype; prepare your audience. Be [E6] Hugh Dubberly and Doris Mitch. © Apple Com-
puter, Inc., 1987. The Knowledge Navigator. (Video.)
clear about what design questions are being ex-
plored with a given prototype—and what are not. [E7] Masamichi Udagawa. © Apple Computer Inc.,
Communicating the specific purposes of a proto- 1995. Project team: Charles Hill, Heiko Sacher,
type to its audience is a critical aspect of its use. It Nancy Silver, Masamichi Udagawa.
is up to the designer to prepare an audience for view-
ing a prototype. Prototypes themselves do not nec- [E8] Charles Hill. © Royal College of Art, London,
1992. Project team: Gillian Crampton Smith, Eleanor
essarily communicate their purpose. It is especially
Curtis, Charles Hill (Royal College of Art), Mike Scaife
important to clarify what is and what is not ad- (Sussex University), and Philip Joe (IDEO, London).
dressed by a prototype when presenting it to any [E9] Tom Bellman, Byron Long, Abba Lustgarten.
audience beyond the immediate design team. University of Toronto, Apple Design Project. © Apple
Computer Inc., 1993.
By focusing on the purpose of the prototype—that
[E10] Apple Design Project. © Apple Computer, Inc.,
is, on what it prototypes—we can make better de-
1992.
cisions about the kinds of prototypes to build. With
a clear purpose for each prototype, we can better

15
[E11] Leo Degen. © Apple Computer Inc., Project Houde, S. (1992). Iterative Design of and Interface for
team: Leo Degen, Stephanie Houde, Michael Mills Easy 3-D Direct Manipulation. Human Factors in
(team leader), David Vronay. Computing Systems. Proc. CHI’92. New York: ACM, pp.
135-142.
[E12] Charles Hill. Doctoral thesis. Project team:
Charles Hill, Henry Weller. © University of London, I.D.Magazine. Apple’s Shared Conceptual Model. The
1993. International Design Magazine: 41st. Annual Design
Review, July-August 1995. USA. pp. 206-207
[E13] Leo Degen. © Apple Computer Inc., 1193.
Project team: Leo Degen, Richard Mander, Gitta Kim, S. (1990). Interdisciplinary Collaboration. The Art
Salomon (team leader), Yin Yin Wong. of Human Computer Interface Design (ed. B. Laurel).
Reading, MA: Addison-Wesley. pp.31-44.
[E14] Daniel Rose. © Apple Computer, Inc., 1993.
Project team: Penny Bauersfeld, Leo Degen, Stephanie Mander, R., Salomon, G., Wong, Y.Y. (1992). A ‘Pile’
Houde, Richard Mander, Ian Small, Gitta Salomon Metaphor for Supporting Casual Organization of
(team leader), Yin Yin Wong Information. Human Factors in Computing Systems:
Proc. CHI’92. New York: ACM, pp. 627-634.
[E15] Charles Hill and Stephen Kamlish. © Royal
College of Art, London, 1992. Project team: Gillian Rose, D.E., Mander, R., Oren, T., Ponceleón, D.B.,
Crampton Smith, Eleanor Curtis, Charles Hill, Stephen Salomon, G., Wong, Y. (1993). Content Awareness in a
Kamlish (Royal College of Art), and Mike Scaife File System Interface: Implementing the ‘Pile’ Metaphor
(Sussex University). for Organizing Information. Research and Development
in Information Retrieval: Proc. SIGIR. Pittsburgh, PA:
6.2 References ACM, pp. 260-269.
Degen, L., Mander, R., Salomon, G. (1992). Working
with Audio: Integrating Personal Tape Recorders and Scaife, M., Curtis, E., Hill, C. (1994). Interdisciplinary
Desktop Computers. Human Factors in Computing Collaboration: a Case Study of Software Development
Systems: Proc. CHI’92. New York: ACM, pp. 413-418. for Fashion Designers. Interacting with Computers, Vol.
6. no. 4, pp. 395-410
Dubberly, H. and Mitch, D. (1987). The Knowledge
Navigator. Apple Computer, Inc. videotape. Schrage, M. (1996). Cultures of Prototyping. Bringing
Design to Software (ed. T. Winograd). USA: ACM Press.
Ehn, P., Kyng, M. (1991). Cardboad Computers: pp. 191-205.
Mocking-it-up or Hands-on the Future. Design at Work:
Cooperative Design of Computer Systems (ed. Wong, Y.Y. (1992). Rough and Ready Prototypes:
Greenbaum, J., and Kyng, M.). Hillsdale, NJ: Lawrence Lessons from Graphic Design. Human Factors in
Erlbaum. pp. 169-195. Computing Systems: Proc. CHI’92, Posters and Short-
Talks, New York: ACM, pp. 83-84.
Erickson, T. (1995). Notes on Design Practice: Stories
and Prototypes as Catalysts for Communication.
“Envisioning Technology: The Scenario as a Framework
for the System Development Life Cycle” (ed. Carroll,
J.). Addison-Wesley.

Hill, C. (1993). Software Design for Interactive

Engineering Simulation. Doctoral Thesis. Imperial
College of Science, Technology and Medicine, Univer-
sity of London.

Hill, C., Crampton Smith, G., Curtis, E., Kamlish, S.

(1993). Designing a Visual Database for Fashion
Designers. Human Factors in Computing Systems:
Adjunct Proc. INTERCHI’93. New York, ACM, pp. 49-
50.

16