Concept blending and dissimilarity:

factors for creative concept generation

process
Yukari Nagai, School of Knowledge Science, Japan Advanced Institute of
Science and Technology, 1-1 Asahidai, Nomi-city, Ishikawa 923-1292, Japan
Toshiharu Taura and Futoshi Mukai, Kobe University, 1-1 Rokkodai, Nada-ku,
Kobe-city 657-8501, Japan

We approached the concept generation stage in design within the framework of

a concept-synthesizing process from two base concepts. We analyzed the
concept generation process by comparing it with the linguistic interpretation
process, from the viewpoints of thought types (property mapping, blending, and
thematic relation) and recognition types (commonalities and alignable and
nonalignable differences). Subjects interpreted a novel nounenoun phrase,
designed a new concept from it, and listed the similarities and dissimilarities
between the nouns. Blending (i.e. generating a new concept that is not included
in the two base concepts but that inherits certain characteristics of the concepts)
and nonalignable differences (i.e. the recognized differences that are unrelated
to the common structure) characterize the creative concept generation process.
Ó 2009 Elsevier Ltd. All rights reserved.

Keywords: generation, design cognition, creativity, conceptual design, design

process

N
umerous studies have investigated the nature of the design process,
and many significant empirical ones have sought to identify its crea-
tive features within a problem solving framework (Cross, 2001).
Cross discovered that the cognitive features of outstanding designers’ thinking
processes were related to problem finding. He categorized these cognitive
processes into taking a broad ‘systems approach’, ‘framing’ the problem and
designing from ‘first principles’ (Cross, 2006) in order to investigate how de-
signers solved problems on the basis of strategic knowledge. Thus, the ‘creative
problem solving’ framework (Holyoak and Thagard, 1995; Ball et al., 2004)
has given rise to rich arguments on the features distinguishing the design activ-
ities of expert and novice designers (Casakin and Goldschmidt, 1999). These
studies have highlighted the concept of analogizing by ‘structure mapping’
(Gentner, 1989), in other words, ‘mental leap’ (Holyoak and Thagard,
1995), as an important factor that enhances solution finding (‘creative’ solu-
Corresponding author: tion by Dorst and Cross, 2001; ‘creative design solution’ by Gero, 1994) in
Yukari Nagai
the design process. Ball et al distinguished between ‘schema-driven analogizing
ynagai@jaist.ac.jp
www.elsevier.com/locate/destud
0142-694X $ - see front matter Design Studies 30 (2009) 648e675
doi:10.1016/j.destud.2009.05.004 648
Ó 2009 Elsevier Ltd. All rights reserved.
 Design process
Figure 1 A design process Source Goal
generation evaluation
model comprising generation
push pull
and problem solving
perspectives Concept generation perspective Problem-solving perspective

(i.e. the recognition and application of abstract experiential knowledge that

could afford a design solution to a familiar problem type) and case-driven
analogizing (i.e. the invocation of a concrete prior design problem whose so-
lution elements could be mapped onto the current problem)’, and reported
that ‘expert designers exhibit more schema-driven analogizing than case-
driven analogizing, whilst novices show the reverse pattern of analogy use’.
Their findings concerning schematic reasoning supported the assumption of
work on analogical relationships between problemesolution in design as
addressed by Visser (1992, 2006) and the differences between the cognitive
processes of experts and novices as addressed in empirical studies of engineer-
ing design. These studies have primarily discussed the role and functions of
analogy in the design process (Christensen and Schunn, 2007). The issues
raised by these studies included the relations between base and target domains
and the types of analogy (i.e. ‘within’ or ‘between’ the domain) in design in
a problem solving framework. However, these views were goal-oriented and
their focus was limited to the problem solving process. Another important as-
pect of the process, namely, ‘generation’, has yet been clarified. Figure 1 illus-
trates the design process, which involves two aspects: the problem solving
aspect, wherein the process begins with the given goal, and the concept gener-
ation aspect, wherein the process begins even in the absence of a goal (Taura
and Nagai, in press).

1 Concept generation stage in creative design

In our study, we focused on the concept generation stage in order to under-
stand the creative features of the design process, as this stage can be considered
when formulating a novel design idea on the basis of the following reasons. As
it is widely known, ‘new ideas evolve from generated ideas that are invoked
from old ideas’ (Finke et al., 1992). A number of design studies have also ad-
dressed the roles of generation in the formation of ideas (namely, concept gen-
eration) related to the original design (Pahl and Beitz, 1984; Liu et al., 2003;
Chiu and Shu, 2007). However, it is possible that the concept generation pro-
cess occurs suddenly in the design process, which makes it difficult to capture it
in the framework of a problem solving process.

Few studies have actually examined the concept generation process. (In this
study, we use the term ‘concept generation’, which, according to Ulrich and Ep-
pinger (2004), includes ‘idea generation’ to represent a formulated design idea
that is a ‘product concept’. The term ‘concept’ is used to represent not only the

Concept blending and dissimilarity: factors for creative concept generation process 649
 image but also the object (natural and artifactual) that is in the person’s mind).
For instance, the mechanism of ‘concept generation’ has been studied to deter-
mine how it leads to the emergence of novel ideas in a framework of the early
stages of design (Liu et al., 2003; Georgiev et al., 2008). Liu et al represented the
process of idea generation as a divergenteconvergent process and described
how designers use multiple levels of abstraction to combine the candidate con-
cepts. Liu et al succeeded in developing a rational concept generation process
model and suggested a method to enhance this process through the use of stim-
uli. More precisely, Chiu and Shu (2007) proposed a methodology that in-
volved enhancing the concept generation process in the design process by
using verbs as stimuli. They investigated how designers use words as stimuli
to enhance concept generation by employing stimuli sets which were organized
in an abstract-concrete hierarchy. They reported interesting phenomena,
namely, that increasing the number of transitive verbs as stimuli may increase
the number of complete concepts the designers generate from their findings,
that words from lower levels of the hierarchy tended to result in a higher per-
centage of complete concepts, and that transitive verbs tended to result in
more success in the development of complete concepts. However, the detail
structures of the concept generation process resulting in the formulation of cre-
ative ideas and its essential factors to generate new ideas have not been identi-
fied. The use of these earlier approaches, which focused on the roles of language
and the effectiveness of using words as stimuli for concept generation, was ev-
ident in design activities such as the ‘word graph’ (Segers et al., 2005).‘Word
graphs’ are used in a computer-aided design system for architectural design,
which projects the word relation graphs that are considered interesting by
the designers. Segers et al. investigated how architect designers’ generation of
design ideas was stimulated when they looked at the structure of words pre-
sented by the word graph system. The word graph probably helped change
the designers’ perspective, helped break their mental fixations, and helped
them avoid making searching errors.

We consider that the above-described methodologies can help designers to

form concepts more quickly (i.e. generate a larger number of concepts, or re-
duce the number of failures), and these concepts serve as appropriate design
ideas in searching for the appropriate candidate concepts by languages. How-
ever, how and why these words led to the creative generation of concepts has
not been explained. Further, how these supports affect the creativity levels of
the outcomes also remains unknown. The identification of the underlying fac-
tors responsible for creative outcomes (generated concepts) requires an under-
standing of the cognitive process (i.e. how designers recognize the given
stimuli) in the concept generation process.

Thus, we focused on the concept generation process from the viewpoint of cre-
ativity. As an extension of our previous study (Taura et al., 2007), this paper
attempts to capture the essence of the concept generation process by (1) we

650 Design Studies Vol 30 No. 6 November 2009

 compare a design task with a non-design task and (2) analyze the essence of
creativity from the viewpoint of the manner in which things are viewed.

With respect to (1), a design task is compared with a linguistic interpretation

task, and with respect to (2), we focus on the dissimilarities between the two
processes.

We adopt the process of synthesizing two base concepts (hereafter called base
concepts), which serves as a framework to capture the concept generation pro-
cess in design. The advantage of this process is that it is the simplest and most
essential process for generating a new concept from existing ones (Rothenberg,
1979; Lubart, 1994). Furthermore, it is suitable because of three reasons,
which are related to empirical, framing, and experimental aspects, respectively.

The first reason is related to the empirical aspect. The concept-synthesizing

process is found in an actual field. Empirically, the invention of the art
knifedthe first snap-off blade cutterdis an appropriate example (Figure 2).
The inspiration for this original idea stemmed from the synthesis of two
concepts, namely, chocolate segments that can be broken off and the sharp
edges of broken glass (Taura et al., 2005).

The second reason is related to the framing aspect. The concept-synthesizing

process involves typical important concept generation processes: property map-
ping, concept blending, and concept integrating. Analogizing by ‘property map-
ping’ is known to play a crucial role in the creative design process (expressed as
‘visual analogy’ by Goldschmidt, 1999). It is also considered to be a concept
creation method involving the transfer of some features from an existing con-
cept to another concept (Gero and Maher, 1993; Gero and Kazakov, 1998).
In practice, it is frequently used in the design process and is regarded as the
most effective design process with respect to the synthesis of two concepts.

Figure 2 Design idea for an

art knife by combining two
conceptsdbroken glass and
chocolate segments

Concept blending and dissimilarity: factors for creative concept generation process 651
 For example, the concept of a ‘white tomato’ can be formed from two individ-
ual concepts, namely, ‘tomato’ and ‘snow’ (Figure 3).

On the other hand, in studies on cognitive linguistics, Fauconnier (1994) ana-

lyzed how conceptual integration develops mental products and the manner in
which one can position the systems of mapping and blending between mental
spaces. He demonstrated that conceptual integration operates on two input
mental spaces to yield a third space, which is termed as ‘the blend’. This
blended space inherits partial structural features from the input spaces and
has emergent structural features of its own (Fauconnier and Turner, 2002).
This concept blending is also a type of concept-synthesizing process. For ex-
ample, from ‘tomato’ and ‘snow’, the concept of ‘powdered ketchup’, which
is used like powdered cheese on a dining table, can be designed.

Further, research that recognizes the relation between two concepts has re-
vealed that there are two types of relationsdtaxonomical and thematicdbetw-
een two concepts (Shoben and Gagne, 1997). The former is a relation that
represents the physical resemblance between two objects, and the latter repre-
sents the relation between two concepts through a thematic scene. In design,
the outcome (hereafter called design product) must be meaningful to people.
Therefore, the designer must carefully consider not only the attributes of the
design product (shape, material, etc.) but also its function and interface; in
other words, consideration of the human factor is important. Consequently,
concept integrationdin which the concepts are synthesized by using the the-
matic relationdis found to play a considerably important role in the creative
design process. With respect to the example of ‘tomato’ and ‘snow’, the con-
cept of a ‘refrigerator that can humidify the food in it’ is designed from the
scene of the situationda tomato stored in snow.

Design Property Concept Concept

processes mapping blending integrating in
thematic
relation

Example e.g. ‘white e.g. e.g.

tomato’for ‘powdered ‘humidifying
snow-tomato ketchup’for refrigerator’
snow-tomato for snow-
tomato

Figure 3 Three types of con-

cept-synthesizing processes
(snowetomato)

652 Design Studies Vol 30 No. 6 November 2009

 As mentioned above, it is found that all the three essential concept generation
processesdproperty mapping, concept blending, and concept integratingdcan
be discussed on the basis of the concept-synthesizing process.

The third reason is related to the experimental aspect. By considering the two
base concepts as a compound phrase composed of two nouns (hereafter called
nounenoun phrase), one can compare the design process with the linguistic in-
terpretation process. In the field of linguistic studies, many results have been
accumulated from the study of nounenoun phrases (Hampton, 1997; Costello
and Keane, 2000). Ward et al (1995) tested how people interpret a nounenoun
phrase (for example, the term ‘computer dog’) and addressed that there was
a divergence of interpretations. Further, they also used this term as a source
of ideas for new inventions by carrying out experiments. They asked students
‘to envision what a computer dog would be like, if it were a variant on a mouse’
and reported that ‘students produced several innovative ideas’.

We consider that the experiment methods proposed by Ward et al can be de-

veloped into a design task for concept generation on the basis of the experi-
ments for the investigation of conceptual design processes (Harakawa et al.,
2005). In addition, the interpretation tasks of a phrase have been examined
conventionally in cognitive linguistic studies. In particular, the interpretation
process of nounenoun phrases has been intensively investigated (Wisniewski,
1996). Therefore, nounenoun phrases can be used as base concepts from
which a new concept is generated, and the phrases can then be interpreted.

In the field of linguistic studies, it is revealed that a novel nounenoun phrase is

interpreted through the following three processes: property mapping, hybrid
linking, and relation linking (Wisniewski, 1996). For example, a knifeefork
can be interpreted as follows: a knife-shaped fork, through the property map-
ping process; one-half as a knife and the other half as a fork, through the hy-
brid linking process; and a knife and fork set used together while eating,
through the relation linking process (Figure 4).

We were able to clarify the three types of linguistic interpretation processes

corresponding to the concept-synthesizing process and categorize them, as
presented in Table 1 (Nagai and Taura, 2006). By using this correspondence,
the design process can be compared with the linguistic interpretation process.
Hereafter, the term ‘blending’ is used for hybrid linking and concept blend-
ing, and ‘thematic relation’ is used for relation linking and concept
integrating.

In our previous experiment, it was revealed that the proportion of property

mapping was lower in the design tasks than in the interpretation tasks. In con-
trast, the proportion of blending was higher in the design tasks than in the in-
terpretation tasks (Taura et al., 2007). This result indicates that the nature of

Concept blending and dissimilarity: factors for creative concept generation process 653
Figure 4 Three types of inter-
pretation processes for the
nounenoun phrase (knifee
fork)

the concept generation process is based on blending. The reason for this is as-
sumed to be as follows. Design products developed by property mapping are
limited in terms of originality, since property mapping cannot extend beyond
the domain of the given concept. In contrast, concept blending can develop
a truly new concept, because the concept developed by this process does not
belong to either domain of the base concepts. Therefore, concept blending is
assumed to characterize the concept generation process, which pursues high
originality. On the other hand, in the interpretation process, the given phrases
are interpreted naturally. Therefore, it is assumed that concept blending is
used more in the design task than in the interpretation task. However, with re-
spect to this assumption, we conducted the experiment only once. Therefore,
this assumption needs to be confirmed by conducting a second and plenary ex-
periment. Further, the mechanism of the blending operation in the design task
needs to be investigated in order to verify this assumption.

Table 1 Classification of the process types for both the linguistic interpretation and design processes

Property mapping Blending Thematic relation

Linguistic interpretation process

Property mapping Hybrid linking Relation linking
(e.g. ‘a knife-shaped fork’ for knifeefork) (e.g. ‘one-half is a knife and (e.g. ‘a knife and fork set’
the other half is a fork’ for knifeefork)
for knifeefork)
Design process
Property mapping Concept blending Concept integration
(e.g. ‘white tomato’ for snowetomato) (e.g. ‘powdered ketchup’ in thematic relation
for snowetomato) (e.g. ‘humidifying refrigerator’
for snowetomato)

654 Design Studies Vol 30 No. 6 November 2009

 On the basis of the result of our previous experiment (Nagai and Taura, 2006),
we also paid attention to thematic relation because it was assumed to be
strongly related to creative outcomes. However, the proportions of thematic
relation in the design task and interpretation task did not show a significant
difference in our next experiments (Taura et al., 2007). Therefore, we chose
the relationships between property mapping and blending in the design task
and interpretation task as the main focus in this study.

In this study, we focus on recognition types (commonalities and alignable and

nonalignable differences). Markman and Wisniewski (1997) explained the no-
tions of alignable and nonalignable differences as follows. ‘Alignable differ-
ences are coded for both references to values along a single dimension, such
as a sled carries more than one person and a ski carries only one person, as
well as for implicit references, such as sleds and skis carry different number
of people. Nonalignable differences are coded for all other differences that
were listed. These differences simply focused on a disparity between the two
items without highlighting a common dimension. An example of a nonaligna-
ble difference would be that an airplane is solid but a puddle is not’. Further, it
was reported that more commonalities and alignable differences were listed for
similar pairs than for dissimilar pairs, while more nonalignable differences
were listed for dissimilar pairs than for similar pairs (Markman and Wisniew-
ski, 1997; Wilkenfeld and Ward, 2001).

Now, we focus on the recognition types in the concept-synthesizing process,

first taking property mapping into consideration. Property mapping is consid-
ered to involve the transfer of some features from an existing concept to an-
other concept. Therefore, the feature recognized in property mapping is
assumed to be an alignable difference since, in property mapping, the feature
recognized in the existing concept displaces the corresponding feature in
another concept. This displacement implies that both these features involve
different values along a single dimension. For example, ‘white tomato’ in
Figure 3 is obtained by transferring the feature of ‘white’ to ‘tomato’. Here,
the recognized feature ‘white’ is classified as an alignable difference, since
‘white’ is the value of color and tomato has another value of color, i.e. ‘red’.
On the other hand, in concept blending, the features recognized in the two syn-
thesized concepts need not be alignable, since these two features are blended to
yield a new concept. For example, in ‘powdered ketchup’ in Figure 3, the rec-
ognized feature ‘powder’ is classified as a nonalignable difference, since the
corresponding feature of ‘powder’ is thought to be non-recognizable in
‘tomato’. Therefore, the nonalignable difference is assumed to be related to
concept blending in the design task.

Further, in our previous study, it was found that if the base concepts are very
dissimilar, a highly creative design product may be obtained (Taura et al.,
2005). By reconsidering this finding from the viewpoint of the recognition

Concept blending and dissimilarity: factors for creative concept generation process 655
 types, we can assume that the creativity in concept blending is related with
recognizing the base concepts as those having nonalignable differences.

Based on the above consideration, we constructed the following hypotheses:

1. The concept blending process characterizes the concept generation

process.
2. Nonalignable differences are related to concept blending and creativity in
the concept generation process.

Although we clarified the two hypotheses as given above, the following

questions remain unanswered.

First, is the recognition process of the nonalignable differences manifested during

the concept generation process or is it an inherent trait? This is a very interesting
question from the viewpoint of learning or teaching the design.

Second, what is the causal relation between the nonalignable differences and
blending? Which one is the cause of the other?

The first question is investigated in the experiment and the second is discussed
at the end of this paper.

We considered that the definition of creativity lies in the generation process,

which leads to the creative outcomes of the design product. We have estimated
creativity by focusing on the relationships that exist between the generation
process and creativity of the generated concept (outcome or ‘design product’).

2 Outline of the experiment

In the experiment (Nagai et al, 2008), the subjects were required to perform the
following three tasks: interpret a novel nounenoun phrase (interpretation
task), design a new concept from the same phrase (design task), and list the
similarities and dissimilarities between the two nouns (similarity and dissimi-
larity listing task). Prior to the experiment, we conducted a preliminary exper-
iment in order to select the nounenoun phrases to be used in the main
experiment. The first and second tasks in the main experiment were conducted
in order to verify the two hypotheses. The third task was conducted in order to
answer the first question.

The responses obtained were analyzed from the viewpoint of the thought types
(property mapping, blending, and thematic relation) and recognition types
(commonalities and alignable and nonalignable differences). Further, the cre-
ativity in the design products was analyzed as follows. First, the design prod-
ucts were evaluated from the viewpoints of originality and practicality.
Second, the features which were enumerated by explaining the design products

656 Design Studies Vol 30 No. 6 November 2009

 and the responses to the interpretation task were judged for whether or not
they were emergent features.

2.1 Interpretation task

The interpretation task consisted of two sub-tasks. First, the subjects were
asked to naturally interpret the nounenoun phrases (the ‘interpretation
task’). Second, they were required to enumerate some words (the ‘interpreta-
tion feature’) to explain each interpretation (the ‘interpretation feature enu-
merating task’). The responses to the interpretation task were analyzed from
the viewpoint of the thought types. The responses to the interpretation feature
enumerating task were analyzed from the viewpoint of the recognition types
and the emergence of features.

2.2 Design task

The design task also consisted of two sub-tasks. First, the subjects were re-
quired to design a new concept from the nounenoun phrases (the ‘design
task’). They were required to not only draw a sketch of the concept but also
explain the concept by using the terms in a sentence. Second, they were re-
quired to enumerate some words (‘design feature’) to explain the features of
each concept (the ‘design feature enumerating task’). The design products
(hereafter, the term ‘design product’ is used to imply something that involves
not only sketch but also the sentence) are analyzed from the viewpoint of the
thought types and creativity (originality and practicality). The responses to the
design feature enumerating task are analyzed from the viewpoint of the recog-
nition types and the emergence of features.

2.3 Similarity and dissimilarity listing task

In this task, the subjects were required to compare the two nouns of the noune
noun phrase used in the interpretation task (as well as the design task) and to list
the common (similarities) and different features (dissimilarities) (the ‘similarity
and dissimilarity listing task’). The responses to the similarity and dissimilarity
listing task were analyzed from the viewpoint of the recognition types.

3 Experimental method
3.1 Selecting the nounenoun phrases used in the preliminary
experiment
The nounenoun phrases to be used in the preliminary experiment were se-
lected according to the following procedures.

First, for the 1055 words listed in the associative concept dictionary (Ishizaki,
2007), the number of associations for each word was investigated, and the
words whose associations were between 168 and 299 (mean  SD) were
selected in order to control the associative effectiveness (Wilkenfeld and
Ward, 2001) in design; as a result, 698 words were selected. Next, these selected

Concept blending and dissimilarity: factors for creative concept generation process 657
 words were classified into eight categories (furniture, musical instrument,
container, natural item, artificial item, tool, wheeled vehicle, and non-wheeled
vehicle) and exceptions were drawn by referring to the method mentioned in
Wilkenfeld and Ward (2001). Finally, 20 nounenoun combination phrases
were selected at random such that the two nouns of each phrase did not belong
to the same category. These 20 nounenoun combination phrases were selected
for use in the preliminary experiment.

3.2 Preliminary experiment for selecting nounenoun phrases

used in the main experiment
In the preliminary experiment, 18 subjects were asked to compare two words
and list the common (similarities) and different features (dissimilarities) be-
tween the two. We planned to select the nounenoun phrases such that the
number of listed common and different features was approximately the
same, and the variance was large; this was done according to the following
guidelines:

 The difference between the mean of the number of common features and
that of different features is lower than the average (0.6), which is obtained
by calculating the average of the differences between the mean of the num-
ber of common features and that of different features among the responses
by the subjects.
 The standard deviation of the number of common features is higher than
the overall average (1.0), which is obtained by calculating the average of
SD of the common features among the subjects’ responses.
 The standard deviation of the number of different features is higher than
the overall average (1.1), which is obtained by calculating the average of
SD of the difference features among the subjects’ responses.

As a result, the following six nounenoun phrases were chosen: shipebox,

pianoeguitar, deskeelevator, drawereplate, shipeguitar, and bookedesk
(Table 2). These six nounenoun phrases were used in the interpretation task
and the similarity and dissimilarity listing task.

Next, two nounenoun phrases used for the design task were selected accord-
ing to the following guidelines:

 Do not choose nounenoun phrases such that the same noun is included in
both nounenoun phrases.
 Do not choose a nounenoun phrase that can be interpreted as a commonly
known phrase.
 Choose a nounenoun phrase that is suitable for a design task.

As a result, two nounenoun phrasesddeskeelevator and shipeguitardwere

selected.

658 Design Studies Vol 30 No. 6 November 2009

Table 2 Nounenoun phrases used in ‘the interpretation task’ and ‘similarity and dissimilarity listing task’

Word A Word B Category A Category B

Ship Box Non-wheeled vehicle Container

Piano Guitar Musical instrument Musical instrument
Desk Elevator Furniture Non-wheeled vehicle
Drawer Plate Furniture Container
Ship Guitar Non-wheeled vehicle Musical instrument
Book Desk Manufactured item Furniture

3.3 Subjects
The subjects comprised 22 undergraduate and graduate students who were
majors in industrial design. They were divided into two groups, Group A
(11) and Group B (11), in order to control the sequence effect of the tasks
(interpretation task / design task; design task / interpretation task).

3.4 Experimental procedure

The experiment was performed using a booklet that included the task instruc-
tions as well as the answer sheets. This booklet consisted of instructions on
the interpretation task, interpretation feature enumerating task, design
task, design feature enumerating task, and similarity and dissimilarity listing
task. Each group was assigned a different room and was presented with the
tasks. We refrained from providing any oral instructions to ensure that the
subjects realized the existence of two types of booklets, which were to be
used depending on the sequence of the tasks. The experiment was conducted
as follows:

Step 1: Group A performed the interpretation task (1 min for each interpre-
tation; total 6 min), while Group B performed the design task (10 min for
each design; total 20 min).
Step 2: Group A performed the interpretation feature enumerating task (2 min
for each interpretation; total 12 min), while Group B performed the design
feature enumerating task (2 min for each designed concept; total 4 min).
Step 3: Group A performed the design task (10 min for each design; total
20 min), while Group B performed the interpretation task (1 min for each
interpretation; total 6 min).
Step 4: Group A performed the design feature enumerating task (2 min for
each designed concept; total 4 min), while Group B performed the interpreta-
tion feature enumerating task (2 min for each interpretation; total 12 min).
Step 5: Groups A and B performed the similarity and dissimilarity listing task
(2 min for each nounenoun phrase; total 12 min).

In the design task, the subjects were asked to design a new concept; the de-
signed concepts were evaluated on the basis of originality and practicality.

Concept blending and dissimilarity: factors for creative concept generation process 659
 On the other hand, in the interpretation task, they were required to naturally
interpret the given phrases.

4 Method of analysis
The responses obtained in the experiment were analyzed from the viewpoint of
recognition types, thought types, creativity (originality and practicality), and
the emergence of features. In this study, the emergence of the enumerated fea-
tures was analyzed, while the design products were also measured by the eval-
uators from the viewpoint of originality and practicality. In order to accurately
compare the design task with the interpretation task, only the responses to
deskeelevator and shipeguitar, which were used in the design task, were
analyzed.

4.1 Classification of the recognition types

We classified the design features and interpretation features and the responses
to the similarity and dissimilarity listing task on the basis of the recognition
types (commonalities and alignable and nonalignable differences) for the
two nouns of the nounenoun phrase used in the interpretation task (as well
as the design task). This classification was done in accordance with the stan-
dards that were set by us in reference to those listed by Markman and Gentner
(1993a,b). The classification standard and examples are shown in Table 3.

4.2 Classification of the thought types

The design product (sketch and sentence) and the interpretation were classified
on the basis of the thought types according to the classification standard pre-
sented in Table 4; these were formulated and used in our previous study (Taura
et al., 2007) in accordance with Wisniewski (1996). This classification was used
to categorize the design products and the interpretation. Therefore, this clas-
sification is not actually based on the thinking process but on the outcomes.
The classification standard is as follows.

4.2.1 Property mapping

The properties of one noun are mapped onto the other noun.

4.2.2 Blending
The category of blending corresponds to ‘hybrid’. Something can have the
properties of both A and B and yet be neither A nor B.

4.2.3 Thematic relation

The category of thematic relation corresponds to ‘relation linking’. The two
nouns are combined from the viewpoint of a thematic scene.

If the products do not fit into any of these three categories, they are classified
under ‘others’.

660 Design Studies Vol 30 No. 6 November 2009

Table 3 Classification standard of recognition types (commonality, alignable differences, and nonalignable differences)

Classification standard and example

Commonality
When an identified feature refers to the common feature of concept A (or part of concept A) and concept B (or part
of concept B) or is associated with both concepts.
Example: in the comparison between ‘ship’ and ‘guitar’, ‘toy’ was judged as a commonality, since both ‘ship’ and
‘guitar’ can be toys.
Alignable difference
When an identified feature indicates a dimension and the values of each concept are different along the dimension,
whether it is expressed explicitly or implicitly.
Example: In the comparison between ‘piano’ and ‘guitar’, ‘how to play’ was judged as an alignable difference.
Nonalignable difference
When an identified feature refers to a feature associated with only one concept (or part of the concept).
Example: in the comparison between ‘ship’ and ‘box’, ‘vehicle’ was judged as a nonalignable difference.
Other
Cases that do not fall under any of the above three categories.
Example: in the comparison between ‘ship’ and ‘guitar’, the ‘planter’ was judged as a feature that does not fit into
any category.

4.3 Creativity evaluation

The creativity of the design products (sketch and sentence) were evaluated
from the viewpoint of practicality (whether the idea seemed achievable and
feasible) and originality (whether the idea was innovative and novel), based

Table 4 Classification standard of the thought types (property mapping, blending, and thematic relation)

Classification standard and example

Property mapping
When the response is a type of concept B (A) similar to concept A (B).
When a part of the property (shape) of concept A (B) or the concepts associated with concept A (B) is transferred
into concept B (A).
Example: In the design task of ‘shipeguitar’, ‘ship-shaped guitar’ was judged as property mapping.
Blending
When the response has the properties of both concepts A and B, and it is neither concept A nor concept B.
When the response is related to concept A (B) from the viewpoint of the material or the response is a part of
concept A (B), and it has the property of concept B (A).
Example: In the interpretation task of ‘pianoeguitar’, the ‘thing that is made up of clavier and strings’ was judged
as blending.
Thematic relation
When the response stems from a situation in which concepts A and B are related to each other (e.g. A move to B).
When the response is a type of concept B (A) that is made of concept A (B).
When the response is a type of concept B (A) that is also meaningful with regard to concept A (B).
Example: In the design task of ‘shipeguitar’, ‘the guitar that plays well even on the moving ship’ was judged as
a thematic relation.
Other
Cases that do not fall under any of the above three categories.
Example: In the design task of ‘shipebox’, ‘ship’ is judged as one that does not fit into any category.

Concept blending and dissimilarity: factors for creative concept generation process 661
Figure 5 An example of the
responses

662 Design Studies Vol 30 No. 6 November 2009

 Figure 5 (continued)

on the creativity evaluation given in Finke et al. (1992). Eleven raters evaluated
all the design products on the basis of a five-point scale (1: low and 5: high).
The rating scores were averaged for each design product. The design products
with lower average scores for practicality than the overall average score for
practicality were excluded from the creativity evaluation. For the remaining
design products, the average scores for originality were considered as the mea-
sure of creativity.

P. M. Blending Thematic

25 4 15
Interpretation
Task

16 15 10
Figure 6 Classification of the
responses according to the Design Task

thought types (P. M. means

property mapping) 0% 20% 40% 60% 80% 100%

Concept blending and dissimilarity: factors for creative concept generation process 663
Table 5 Result of the residual analysis on the classification of the responses according to the thought types

Thought type Property mapping Blending Thematic relation

Interpretation task 1.64 3.04** 0.98

Design task 1.64 3.04** 0.98

jresidualj > 1.65 / yp < 0.10; jresidualj > 1.96 / *p < 0.05; jresidualj > 2.58 / **p < 0.01.

4.4 Judgement of emergent features

By referring to Wilkenfeld and Ward (2001), the enumerated features (inter-
pretation features and design features) were judged in terms of whether or
not they were emergent features. When the feature was not found to be an as-
sociative concept of the two nouns (on the basis of which the interpretation
and design tasks were conducted), it was judged as an emergent feature. The
associative concept dictionary (Ishizaki, 2007) and synonym dictionary (Ya-
maguchi, 2006) were used for this judgement. When a feature was found to
be an associative concept of the two nouns in the associative concept dictio-
nary, it was judged as a non-emergent feature. This was done for each feature.
Furthermore, we investigated the synonyms of the associative concepts by us-
ing the synonym dictionary. When the feature was found to be a synonym of
the associative concepts of the nouns, it was also judged as a non-emergent
feature.

5 Results
Seven responses (three for the design task, three for the design feature enumer-
ating task, and one for the interpretation feature enumerating task) were ex-
cluded from the analysis because they were inadequate. First, we examined
the influence of the sequence of the tasks. The results of a chi-square test
were as follows. The proportion of the thought types of Groups A and B
did not display a significant difference. For the interpretation and design tasks,
the chi-square values were 0.96, n.s. and 0.24, n.s., respectively.

An example of the responses is shown in Figure 5.

Alignable Nonalignable
Commonality
difference difference

78 9 74
Interpretation
Task

74 7 109

Design Task
Figure 7 Classification of the
responses according to the
recognition types 0% 20% 40% 60% 80% 100%

664 Design Studies Vol 30 No. 6 November 2009

Table 6 Result of the residual analysis on the classification of the responses according to the recognition types

Recognition type Commonality Alignable difference Nonalignable difference

Interpretation task 1.79y 0.85 2.13*

Design task 1.79y 0.85 2.13*

jresidualj > 1.65 / yp < 0.10; jresidualj > 1.96 / *p < 0.05; jresidualj > 2.58 / **p < 0.01.

5.1 Comparison of the design and interpretation tasks from

the viewpoint of the thought types
The classification of the design products (sketch and sentence) and the inter-
pretation on the basis of the thought types are illustrated in Figure 6. We
found a high proportion of blending in the design products as opposed to
the interpretation. This result corresponds to that of our preliminary experi-
ment (Taura et al., 2007) and reinforces the adequacy of hypothesis (1). The
chi-square test detected a significant difference in the proportion of the
thought types between the two task types (c2 (2) ¼ 9.24, p < 0.01). The result
of the residual analysis indicated a significant difference only in blending as
shown in Table 5.

5.2 Comparison of the design and interpretation tasks from

the viewpoint of the recognition types (commonalities and
alignable and nonalignable differences)
According to the standard presented in Table 3, the interpretation features and
design features were classified on the basis of the recognition types.

The results are illustrated in Figure 7. In the chi-square test, a significant dif-
ference was detected in the proportion of the recognition types between the in-
terpretation features and design features (c2 (2) ¼ 4.69, p < 0.10). The result of
the residual analysis indicated that the proportion of nonalignable differences
in the design features was higher than that in the interpretation feature, while
the proportion of commonalities was low. It is assumed that more attention is

Table 7 Mean of the proportion of recognition types among the responses classified into each thought type for interpretation
task/design task (based on the feature enumerating task) (P.M. means property mapping)

P.M. Blending Thematic relation

Commonalities
Interpretation 0.471 0.243 0.497
Design product 0.448 0.314 0.504
Alignable differences
Interpretation 0.052 0.125 0.141
Design product 0.028 0.011 0.053
Nonalignable differences
Interpretation 0.477 0.632 0.362
Design product 0.524 0.675 0.443

Concept blending and dissimilarity: factors for creative concept generation process 665
 Interpretation Task Design Task
1.0
Figure 8 Mean of the propor-

Nonalignable Differences
tion of nonalignable differ- 0.8

Proportion of
ences in the thought types
for the interpretation task/de- 0.6

sign task (feature enumerat-

0.4
ing task) (property mapping
expresses as P.M.). Note: 0.2

The error bar shows the stan-

0.0
dard error of the mean P. M. Blending Thematic

paid to nonalignable differences in the design process than in the interpretation

process as shown in Table 6. This result is consistent with hypothesis (2).

5.3 Comparison of the thought and recognition types

First, with respect to the interpretation features and design features, we deter-
mined the proportion of the recognition types (commonalities and alignable
and nonalignable differences) for each interpretation and design product. Fur-
ther, we calculated the average of the proportions of the design products and
interpretations classified under each thought type (property mapping, blend-
ing, and thematic relation); this result is presented in Table 7. A two-factor fac-
torial ANOVA (between-subjects factorial design; factor 1: interpretation or
design task, factor 2: thought types) indicated a significant difference only in
the factor of the thought type with respect to the proportion of the nonalign-
able differences (F (2,76) ¼ 3.22, p < 0.05). This suggests that the thought types
may be characterized by nonalignable differences (Figure 8).

In the analysis in the two preceding sections, it was found that blending and
nonalignable differences characterize the design process. Confirming this find-
ing, the result obtained above (Figure 8) suggests that nonalignable differences
are related to blending.

Table 8 Mean of the proportion of recognition type among the responses classified into each thought type for interpretation
task/design task (based on the similarity and dissimilarity listing task) (P.M. means property mapping)

P.M. Blending Thematic relation

Commonalities
Interpretation 0.392 0.775 0.41
Design product 0.39 0.451 0.405
Alignable differences
Interpretation 0.205 0.133 0.186
Design product 0.249 0.163 0.117
Nonalignable differences
Interpretation 0.403 0.092 0.405
Design product 0.361 0.386 0.478

666 Design Studies Vol 30 No. 6 November 2009

 Interpretation Task Design Task
Figure 9 Mean of the propor-
1.0

Nonalignable Differences
tion of nonalignable differ-
ences in the thought type for 0.8

Proportion of
the interpretation task/design
0.6
task (similarity and dissimi-
larity listing task) (property 0.4
mapping expressed as
P.M.). Note: The error bar 0.2

shows the standard error of

0.0
the mean P. M. Blending Thematic

Second, with regard to the responses obtained in the similarity and dissimilarity
listing task, we determined the proportion of the recognition types (common-
alities and alignable and nonalignable differences). Further, we calculated the
average of the proportions of the responses belonging to each thought type
(property mapping, blending, and thematic relation); the thought types were
determined on the basis of the responses corresponding to the thought types
of the design product and interpretation with respect to the same nounenoun
phrase used in the similarity and dissimilarity listing task. It is assumed that this
average indicates the manner in which things or concepts are viewed by the sub-
jects, whose design products and interpretations are classified under each
thought type. This result is presented in Table 8. A two-factor factorial
ANOVA revealed that there was no significant difference in the factors.

This result suggests that focusing on nonalignable differences is not an inher-

ent trait of the subjects (Figure 9); rather, it occurs during the design and
interpretation processes.

5.4 Comparison of the design and interpretation tasks from

the viewpoint of the emergence of the features
The mean of the emergent features (interpretation features and design fea-
tures), which were judged according to the standard presented in Section

5.0
Emergent Features

4.0

3.0

Figure 10 Mean of the num-

2.0
ber of emergent features.
Note: The error bar shows 1.0

the standard error of the

0.0
mean Interpretation Task Design Task

Concept blending and dissimilarity: factors for creative concept generation process 667
 5.0
4.5
4.0
3.5

Originality
3.0
2.5
2.0
1.5
1.0
Figure 11 Relation between
0.5
score of originality and pro-
0.0
portion of nonalignable differ- 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
ences in blending Proportion of nonalignable differences

4.4, is illustrated in Figure 10. This figure shows that more emergent features
were used for explaining the design product rather than for explaining the in-
terpretation (two-sided test: t (82) ¼ 2.36, p < 0.05). This result indicates that
more novel features emerge during the design process rather than during the
interpretation process.

5.5 Relation between creativity and recognition types

The creativity of the design product is evaluated according to the procedure
determined in Section 4.3. Kendall’s coefficient of concordance shows the sig-
nificant coincident factor in both originality and practicality (originality:
W ¼ 0.34, c2 (40) ¼ 148.86, p < 0.01; practicality: W ¼ 0.32, c2 (40) ¼ 142.18,
p < 0.01). Therefore, this evaluation result was used for the following analysis.
The remaining design products with higher average scores for practicality than
the overall average score for practicality are 9 (property mapping), 6 (blend-
ing), and 4 (thematic relation).

No correlation between the score of originality and the proportion of recogni-

tion type was detected (commonality and alignable and nonalignable

5.0
4.5
4.0
3.5
Originality

3.0
2.5
2.0
1.5 y = -0.095x2 + 0.8549x + 1.6245
1.0 R = 0.68
0.5 (p < .01)
Figure 12 Relation between
0.0
score of originality and num- 0 2 4 6 8 10
ber of emergent features Number of emergent features

668 Design Studies Vol 30 No. 6 November 2009

 5.0
4.5
4.0
3.5

Originality
3.0
2.5
2.0
1.5 y = -0.1004x2 + 0.8617x + 1.893
1.0 R = 0.82
Figure 13 Relation between
0.5 (p < .01)
score of originality and num-
0.0
ber of emergent features in 0 2 4 6 8 10
property mapping Number of emergent features

differences) for all the design products. However, a strong correlation was de-
tected between the score of originality and the proportion of the commonality
and nonalignable differences for the design products classified into blending
(nonalignable difference: r ¼ 0.80, F (1,4) ¼ 7.11, 0.05 < p < 0.10; commonal-
ity: r ¼ 0.80, F (1,4) ¼ 7.11, 0.05 < p < 0.10) (Figure 11).

This result indicates that focusing on nonalignable differences is related with

originality in concept blending, which characterizes the design process. This
result is consistent with hypothesis (2).

5.6 Relation between creativity and the emergence of features

The relation between the number of emergent features and the score of origi-
nality is illustrated in Figures 12e14. A regression analysis detected a signifi-
cant curve regression (R ¼ 0.68, p < 0.01) rather than a linear regression.
Furthermore, a regression analysis for the design products classified into prop-
erty mapping and blending detected a stronger significant regression (property
mapping: R ¼ 0.82, p < 0.01; blending: R ¼ 0.94, p < 0.05), as opposed to the
design products classified into thematic relation. This result indicates that

5.0
4.5
4.0
3.5
Originality

3.0
2.5
2.0
1.5
y = -0.2727x2 + 2.2727x - 0.9545
1.0
Figure 14 Relation between R = 0.94
0.5 (p < .05)
score of originality and num-
0.0
ber of emergent features in 0 2 4 6 8 10
blending Number of emergent features

Concept blending and dissimilarity: factors for creative concept generation process 669
 there exists an appropriate emergent level for inducing high originality in
design.

6 Discussion
This experiment revealed that the concept generation process is characterized
by concept blending and the recognition of nonalignable differences. With re-
spect to the relationship between these two factors, we assume the following
models:

1. First, the designer captures the features of the two nouns (base concepts)
from the viewpoint of nonalignable differences and then adopts the con-
cept blending process so that the nonalignable features are used.
2. First, the designer attempts to adopt the concept blending process and
then captures the features of the two nouns (base concepts) from the view-
point of nonalignable differences so that the concepts can be blended.

We propose that both the processes can coexist, and which process is adopted
depends on the condition in which the designer is. This issue, however, requires
further investigation.

We now proceed to discuss findings concerning various aspects of our exper-

iment in light of previous work on important issues in the field.

6.1 Cognitive features in concept generation

We discussed the creative features of the design process, particularly in the
concept generation process, by focusing on the relationships between the gen-
eration process and the outcomes (generated ideas for design products), which
we evaluated on the basis of not only practicality but also originality. Based
on the results of the relation between creativity and recognition type, we con-
sider that the originality of the outcomes was affected by the type of recogni-
tion of nonalignable differences particularly in the case of concept blending,
which characterizes the design process. This result reveals the possibility
that the difference between the features of the two base concepts evoked the
designers’ creativity, which supports the results of the experiments of our pre-
vious study (2005), wherein designers obtained high originality scores on con-
cept synthesis of base concept pairs that had a high degree of difference
compared to the case in which the pairs had moderate or low levels of differ-
ence between them. This finding appears to be in accordance with Wilkenfeld
and Ward’s (2001) results, according to which the emergence results of inter-
pretation tasks involving nounenoun combinations were high only in the case
of dissimilar pairs.

6.2 Language stimuli for design

The findings of this research can be discussed by considering the findings of
earlier studies on stimuli for concept generation. Previous research on this

670 Design Studies Vol 30 No. 6 November 2009

 topic has attempted to find a means of enhancing such concept generation
(Chiu and Shu, 2007). Significant earlier studies have contributed to the devel-
opment of methods that can be used to support creative design by using lan-
guage stimuli for idea generation by using tools such as the ‘word graphs’
(Segers et al., 2005). Although our findings do not directly explain why stimuli
enhance concept generation in creative design, they can help us gain an under-
standing of the underlying cognitive process of concept generation.

To consider the roles of stimuli in enhancing creativity in design, both studies

(Segers et al., 2005; Chiu and Shu, 2007) presented language that was external
to the designer; in other words, language motivated the designers externally.
In contrast, we investigated the inner processes of the subjects. The subjects
had to perform a start design task by using a nounenoun combination and
had to explore the ideas by themselves without any assignment goals. It is
possible that the design tasks themselvesdas internal stimulidmotivated
the subjects’ concept generation. However, in our experiments, in the language
interpretation tasks, the subjects were told ‘to naturally interpret’, and in the
design tasks, they were told to ‘design a new concept’. It is possible that the
dual instructions affected the subjects. We cannot confirm precisely which
wordd‘design’ or ‘new’daffected the designers’ cognitions in the design tasks
and which from among ‘naturally’ and ‘interpret’ affected the designers’ cog-
nition in the language interpretation tasks.

In order to obtain a detailed explanation for why there exist different types of
thoughts and recognition in design tasks and language interpretation tasks,
future studies should carry out another analysis with the designers.

There exists another consideration with regard to language stimuli. Based on

Gagne (2000), Franks and Rigby (2005) argued about the differences in inter-
pretation by comparing relation-based and property-based concept combina-
tions by conducting experiments focusing on gender differences in perception.
Their results suggested that there are differences between male perceptions and
female perceptions. Their study contributes to the discussion on the phenom-
ena of language evolution and helps in understanding the nature of communi-
cation. They investigated the creative interpretation of combined words in
males and females. There were no remarkable relationships between the results
of their study and ours, because we conducted a comparison between interpre-
tation tasks and design tasks. Although we cannot compare the results di-
rectly, we focused on the effects on the designers’ motivation. Franks and
Rigby aimed to investigate gender differences in the creative interpretation
of communication. We consider their study revealed that the subjects were ex-
ternally motivated. In contrast, our results suggest that the subjects were inter-
nally motivated by the tasks. The implementation of our result is design
activity drives the creative process, as follows.

Concept blending and dissimilarity: factors for creative concept generation process 671
 6.3 Motivations of design
Our findings are potentially important in terms of explaining how designers
create a novel concept from existing ones. We consider that difference go be-
yond being merely the opposite of similarity. The results suggest that the ability
to recognize the nonalignable differences between existing entities or attributes
enables designers to generate ideas. In other words, it is important to enhance
the designers’ (or design students’) awareness in order to equip them with the
necessary skills for creating highly original products, by activating their abili-
ties to recognize both commonalities and differences between existing objects.

Through this study, we captured the phenomena of the design process from the
perspectives of thought types and recognition processes arising in the concept
generation stage; however, there are two issues that need to be resolved in or-
der to understand the nature of the design process. The first issue is that the
influence of the two stimuli (tasks of interpretation and design) is yet to be clar-
ified. Second, the cognitive aspect concerning the relationship between
thought type and recognition type (the question of whether the feature is em-
bedded innately or is something which is acquired later) is also a matter that
needs to be discussed.

At the outset of this study, we expressed our views on the concept generation
stage in the design process and proposed relevant arguments supporting the
fact that the concept generation process and the problem solving process
were distinct (see Figure 1). However, these two stages are actually a part of
one sequential process (Taura and Nagai, 2009). To conceptualize the actual
designing process, we need to present an argument based on research on an
entire design process. The findings of this study could perhaps help increase
our knowledge of the design process by considering a generation stage, that
is, a stage involving the inner motivation of a designer. This may help us to
determine what factors drive the design process on the basis of a view that
is different from the previous views emphasizing designers’ external motiva-
tions, that is, a future goal.

7 Conclusion
In this study, we analyzed the characteristics of the concept generation process
in the design process in comparison with the interpretation process. In partic-
ular, we analyzed the characteristics from the viewpoint of the thought types
(property mapping, blending, and thematic relation) and recognition types
(commonalities and alignable and nonalignable differences). Based on the
analysis, it was found that blending and nonalignable differences characterize
the creative design process. In addition, it was suggested that a focus on non-
alignable differences is associated with creativity in the blending process. This
study elucidates the specific characteristics of the design process and other
pertinent issues, with the aim of clarifying on creativity in design.

672 Design Studies Vol 30 No. 6 November 2009

 References
Ball, L J, Ormerod, T C and Morley, N J (2004) Spontaneous analogising in
engineering design: a comparative analysis of experts and novices, Design Studies
Vol 25 No 5 pp 495e508
Casakin, H and Goldschmidt, G (1999) Expertise and the use of visual analogy:
Implications for design education, Design Studies Vol 20 No 2 pp 153e175
Chiu, I and Shu, L H (2007) Using language as related stimuli for concept gener-
ation [Article], Ai Edam-Artificial Intelligence for Engineering Design Analysis and
Manufacturing Vol 21 No 2 pp 103e121
Christensen, B T and Schunn, C D (2007) The relationship of analogical distance to
analogical function and preinventive structure: the case of engineering design,
Memory & Cognition Vol 35 No 1 pp 29e38
Costello, F J and Keane, M T (2000) Efficient creativity: constraint-guided concep-
tual combination, Cognitive Science Vol 24 No 2 pp 299e349
Cross, N (2001) Strategic knowledge exercised by outstanding designers in
Proceedings of Strategic Knowledge and Concept Formation III pp 17e30
Cross, N (2006) Designerly ways of knowing Birkhauser, Basel, Switzerland
Dorst, K and Cross, N (2001) Creativity in the design process: co-evolution of
problem solution, Design Studies Vol 22 No 5 pp 425e437
Franks, B and Rigby, K (2005) Deception and mate selection: some implications
for relevance and the evolution of language in M Tallermann (ed) Language
origins, Oxford University Press, Oxford
Fauconnier, G (1994) Mental spaces: aspects of meaning construction in natural
language Cambridge University Press, Cambridge, UK
Fauconnier, G and Turner, M (2002) The way we think Basic Book, New York,
USA
Finke, R A, Ward, T B and Smith, S M (1992) Creative cognition: theory, research,
and applications MIT Press, Cambridge, USA
Gagne, C L (2000) Relation-based combinations versus property-based combina-
tions: a test of the CARIN theory and dual-process theory of conceptual-combi-
nation, Journal of Experimental Psychology: Learning, Memory and Cognition
No 27 pp 235e254
Gentner, D (1989) The mechanisms of analogical learning in S Vosniadou and
A Ortony (eds) Similarity and analogical reasoning, Cambridge University Press,
Cambridge, UK
Georgiev, G V, Taura, T, Chakrabarti, A and Nagai, Y (2008) Method of design
through structuring of meanings, ASME International Design Engineering
Technical Conferences (IDETC) and Computers and Information in Engineering
Conference (CIE), New York (CD-ROM 10 pages)
Gero, J S and Maher, M (1993) Modeling creativity and knowledge-based creative
design Hillsdale, New Jersey, USA
Gero, J S (1994) Computational models of creative design processes in T Dartnall
(ed) Artificial intelligence and creativity: an interdisciplinary approach, Kluwer
Academic Publishers, Boston, MA pp 269e282
Gero, J S and Kazakov, V (1998) Using analogy to extend the behaviour state
space in design in J S Gero and M L Maher (eds) Computational models of creative
design IV, University of Sydney, Sydney, Australia pp 113e143
Goldschmidt, G (1999) Visual analogy e a strategy for design reasoning and learn-
ing in C Eastman, M McCracken and W Newstetter (eds) Design knowing and
learning: cognition in design education, Avebury
Hampton, J A (1997) Emergent attributes in combined concepts in S M Smith and
J Vaid (eds) Creative thought pp 83e127

Concept blending and dissimilarity: factors for creative concept generation process 673
 Harakawa, J, Nagai, Y and Taura, T (2005) Study on conceptual synthesis in de-
sign creation e role of thematic relation in creativity in Proceedings of Interna-
tional Design Congress IASDR 2005 (in CD-ROM)
Holyoak, K J and Thagard, P (1995) Mental leaps: analogy in creative thought
MIT Press, MA, USA
Ishizaki, S (2007) Associative concept dictionary (ver. 2) Keio University (in CD-
ROM) (in Japanese)
Liu, Y C, Bligh, T and Chakrabarti, A (2003) Towards an ‘ideal’ approach for con-
cept generation, Design Studies Vol 24 No 4 pp 341e355
Lubart, T (1994) Creativity in R J Stenberg (ed) Thinking and problem solving,
Academic Press, USA pp 289e332
Markman, A B and Gentner, D (1993a) Structural alignment during similarity
comparisons, Cognitive Psychology Vol 25 No 4 pp 431e467
Markman, A B and Gentner, D (1993b) Splitting the differences: a structural align-
ment view of similarity, Journal of Memory and Language Vol 32 pp 517e535
Markman, A B and Wisniewski, E J (1997) Similar and different: the differentia-
tion of basic-level categories, Journal of Experimental Psychology: Language,
Memory & Cognition Vol 35 pp 54e70
Nagai, Y and Taura, T (2006) Formal description of conceptual synthesizing
process for creative design in Proceedings of the 2nd International Conference
on Design Computing and Cognition ’06
Nagai, Y, Taura, T and Mukai, F (2008) Concept blending and dissimilarity: fac-
tors for creative design process e a comparison between the linguistic interpreta-
tion process and design process in Proceedings of the 2008 Design Research Society
International Conference (CD-ROM, Sheffield, England, 22 pages)
Pahl, G and Beitz, W (1984) Engineering design Springer/Design Council, London, UK
Rothenberg, A (1979) The emerging goddess: the creative process in art, science, and
other fields University of Chicago Press, Chicago, USA
Segers, N M, Vries, B D and Achten, H H (2005) Do word graphs stimulate de-
sign? Design Studies Vol 26 pp 625e647
Shoben, E J and Gagne, C L (1997) Thematic relation and the creation of com-
bined concepts in T B Ward, S M Smith and J Vaid (eds) Creative thought, Amer-
ican Psychological Association pp 31e50
Taura, T, Nagai, Y and Tanaka, S (2005) Design space blending in Proceedings of
ICED05. 15th International Conference on Engineering Design, (on CD-ROM)
Taura, T, Nagai, Y, Morita, J and Takeuchi, T (2007) A study on design creative
process focused on concept combination types in comparison with linguistic inter-
pretation process in Proceedings of the 16th International Conference on Engineer-
ing Design ICED07 (CD-ROM) 12 pages
Taura, T and Nagai, Y Design insight e a key to studying design creativity, in
John Gero (ed) Studying design creativity, Springer, in press
Taura, T and Nagai, Y (2009) Design creativity: integration of design insight and
design outsight, what is ‘‘what’s the design’’? Special Issue of Japanese Society for
the Science of Design Vol 16-2 No 62 pp 55e60
Ulrich, K T and Eppinger, S D (2004) Product design and development (3rd edn)
McGraw-Hill/Irwin, NY
Visser, W (1992) Designers’ activities examined at three levels: organization,
strategies and problem-solving processes, Knowledge-Based Systems Vol 5 No 1
pp 92e104
Visser, W (2006) The cognitive artifacts of designing Lawrence Erlbaum Associ-
ates, USA
Ward, T B, Finke, R and Smith, S M (1995) Creativity and the mind: discovering
the genius within Perseus Publishing, USA

674 Design Studies Vol 30 No. 6 November 2009

 Wilkenfeld, M J and Ward, T B (2001) Similarity and emergence in conceptual
combination, Journal of Memory and Language Vol 45 No 1 pp 21e38
Wisniewski, E J (1996) Construal and similarity in conceptual combination,
Journal of Memory and Language Vol 35 No 3 pp 434e453
Yamaguchi T (ed) (2006) Japanese thesaurus dictionary, Taisyukan, Tokyo
(in Japanese)

Concept blending and dissimilarity: factors for creative concept generation process 675