Industrial

Ergonomics
ELSEVIER International Journal of Industrial Ergonomics 19 (1997) 105-114

Application studies to car interior of Kansei engineering

Tomio Jindo *, Kiyomi Hirasago
Vehicle Research Laboratory, Nissan Research Center. Nissan Motor Co., Lid. 1. Natsushima-cho, Yokosuka 237. ,lapan

Abstract

This paper describes studies of styling or design specification of passenger car interiors as examples of application of
Kansei engineering, especially regarding the speedometer and steering wheel of a passenger car. These units require easy
operation or good visibility, but we did not consider these functions here and treated only the styling impression. Subjective
evaluations were carried out by semantic differential methods, then analyzed by using multivariate analyses. We gathered the
results concerning the relationship between an impression and characteristics of styling to understand the conditions which
create a desired impression.

Relevance to industry

These studies were carried out to improve the cabins of passenger cars. We can also apply similar methods of styling to
other industry products.

Kevwords: Kansei engineering: SD method; Automobile design: Multivariate analysis

1. Introduction chart. For example 'sporty' is the input and this

system shows the suitable graphics for that adjective
According to a recent passenger car's improve- on the CRT.
ment of basic functions, the user's demand for a car However, this system included many car interior
shifted from functional aspects to a total ambience parts like seats, a speedometer, a steering wheel,
including styling. Therefore, when designing a car switches and so on. Detailed information about each
interior, Kansei engineering is used to grasp vague part's design elements could not be gained because it
demands of the consumer, and develop the car based was impossible to carry out actual experiments using
on the user's words. As an example, there is support so many samples in order to analyze detail design
for expert systems relating styling to car interiors. elements of each part.
This was based upon the analysis data showing a Therefore, we studied only one interior unit with
relationship between human impression and interior detail design elements and analyzed the relationships
design. Fig. 1 shows a rough construction and flow between design elements and impressions. By doing
this, more practical knowledge was gained than by
conventional experiments which study many interior
:' Corresponding author. Tel.: +81 468-67-5158, Fax: +81 units. Adding this data to conventional data, regard-
468-65-5699, E-mail: [email protected]. ing impressions of these units in styling, supports

0169-8141/96/$15.00 Copyright © 1996 Elsevier Science B,V. All rights reserved.

PII S 0 1 6 9 - 8 1 4 1 ( 9 6 ) 0 0 0 0 7 - 8
106 T. Jindo, K. Hirasago / International Journal o['Industrial Ergonomics 19 (1997) 105 114

Input of evaluationadjectives [ Knowledgedatabase digital ones will not be used widely which is why
Inference they are not considered in this paper.
Searchfor corresp?ndingadjectives ~ A~tive d ~ A survey was made of existing analog meter
T
Determinationof designelements clusters and their constituent design elements were
extracted and classified. Typical design elements are
Checkfor contradictions listed in Table 1. A series of subjective evaluations
Determinationof interiorparts were made of those elements. This paper describes
~daGraphic,
k imag~ the results of subjective evaluations and subsequent
--. database.....J analyses that were made of speedometers alone,
Fig. 1. Construction and flowchart of car interior styling support meter cluster layouts, and meter cluster assemblies
system. using photographs of actual meter clusters.

expert systems about car interiors. Car interior styling 2. l. 1. Subjective eualuation of speedometers in isola-
support systems which also have detailed data of tion
design elements will become possible in the future. The design elements of the speedometer that were
chosen for analysis were the scale, lettering, types of
indicators and starting point of the indicator. The
2. Overview of research work types of design elements examined are shown in
Table 2. Using different combinations of these de-
2.1. A study of speedometer on Kansei engineering sign elements, 24 speedometer samples were created
by computer graphics for subjective evaluation. Fig.
The scope of this research was limited to analog 2 shows an example of one of the samples. Eight
speedometers. Digital ones have too many degrees of pairs of adjectives thought to describe typical im-
freedom in styling. They are too complicated to pressions of the speedometer were used by 23 sub-
study and analyze. The number of actual digital jects, 14 men and 9 women, in making evaluations.
speedometers is small, so the result of a study on The subjects assigned a numerical value to each of

Table 1
Design elements of analog meter

1. Meter layout 5. Outside scale, '~\

inside scale Q!J 2
2. Meter types and number Outside scale insidescale
Speedometer
Tachometer 6. Starting point
Fuel level gauge
Water Ievel gauge
etc.
7.Scale type ~t'l L~J~ ~0 40 > 4o
3. Panel color and material 20 40 2{) 40 20 40
Plastic I ~ J L~_! krcrt2Lzz]
Wood
Leather 8.Number s ~ ~
4. Meter shape orientation < e

9. Lettering ~ ~1
',.j Horizontal Centrifugal
Round Semicircular Quarter Ova[ 10. Indicator shape
 12 Jindo, K. Hirasago / lnternational Journal of Industrial Ergonomics 19 (1997J 105-I 14 107

the eight adjective pairs according to a 7-point se-

mantic differential (SD) scale.
A factor analysis was then performed on the
scores assigned to each evaluation adjective pair, and
two factors were extracted. The evaluation adjectives
were then positioned on the two factor axes in terms
of their factor loading. The resulting arrangement of
the adjectives is shown in Fig. 3. One of the factor
axes in the figure is seen to indicate a feeling of
Fig. 2. Exampleof speedometersample.
being easy to understand, represented by the adjec-
tive 'clean-looking'. The other axis is interpreted as
indicating a design factor, represented by the adjec-
tive 'luxurious', In addition, the adjective 'likable' is strong effect on increasing the perceived feeling of
positioned where both the easy-to-understand and luxury. The type-3 lettering in particular was effec-
design factors have positive values. These results tive in projecting a more luxurious impression. Simi-
suggest that easy-to-understand and the design were larly, the scale had a strong influence on the clean-
the two major impressions on which the subjects looking feeling. The type-3 scale with large gradua-
based their evaluation of the speedometer samples. tions was noticeably effective in heightening the
Thus, it can be assumed that an instrumentation image of being clean-looking.
design which satisfies these two aspects should be
favorably received. 2.1.2. S u b j e c t i u e e u a l u a t i o n s o f m e t e r c l u s t e r l a y o u t
A multiple regression analysis technique known Meter cluster samples having two to seven meters
as the Quantification I method was then used to were prepared and evaluated for the purpose of
analyze the relationships between the subjective examining the correlation between the number of
evaluation scores and meter cluster design elements. meters in the same cluster and the evaluation adjec-
This technique is commonly used in Japan to exam- tives. Fig. 4 shows the six types of meter layouts
ine the relationship between quantitative data (the used, which were identical in size to actual meter
scores in this work) and qualitative data (the design clusters. Each meter cluster was given a design
categories of the samples evaluated). The analytical resembling an actual meter assembly like the sample
results obtained for the two adjectives 'luxurious' shown in Fig. 1. The same eight pairs of evaluation
and 'clean-looking' are shown in Table 3. adjectives were used as in the subjective evaluations
The partial correlation coefficients in the table of speedometers. Evaluation scores were assigned
indicate the extent to which each design element according to a 7-point SD scale. Twenty-four sub-
contributes to an explanation of the evaluation adjec- jects, 18 men and 6 women, took part in the evalua-
tive concerned. The bar graph results in the table tions.
show that the lettering of the speedometer had a The evaluation scores were averaged for each
adjective and layout sample, and a factor analysis
was performed on the average scores found for the
Table 2 adjectives. As a result, two factor axes were ex-
Design elements of evaluation test
tracted, as in the subjective evaluations of speedome-
ters. The factor axis interpretations were also identi-
cal, i.e., an easy-to-understand factor and a design
scale type Lettering Indicator shape Starting point factor. The adjectives were arranged on the factor
axes in terms of their factor loading, as shown in
Fig. 5. The meter layout samples have also been
arranged in the figure on the basis of their factor
7 . ... ~ 4 0123 4. <.~---.~-~ ~ ('-"
scores. The following observations can be made
from the results.
108 T. Jindo, K. Hirasago / lnternational Journal of Industrial Ergonomics 19 11997) 105-114

Design axis
2 meters 3 meters

Luxu!i°us•OElegant / @ @ ~ / ~ @ @

Playful • ~etro-looking
• Likable
4 meters 5 meters

oSporty Clean-looking

Easy -to- 6 meters 7 meters

understand
axis
Easy 1o understand
Fig. 3. Arrangement of evaluation adjectives.
Fig. 4. Meter cluster layouts evaluated,
1. Meter clusters with a smaller number of meters
( 2 - 4 ) projected a stronger feeling of being easy
correlation between their perception of the meter
to understand.
cluster assemblies and the design elements. The pho-
2. Meter clusters having an even number of meters
tographs showed ten types of meter cluster assem-
(4 and 6) were evaluated highly in terms of
blies used in Japanese passenger cars. An example is
design and tended to be liked by the subjects.
shown in Fig. 6. Thirteen pairs of adjectives were
3. The 4-meter cluster was rated highly as being
used and scores were assigned according to a 7-point
easy to understand, and the 6-meter cluster was
SD scale.
perceived as being sporty.
The evaluation adjectives consisted of the words
2.1.3. Subjective evaluations using photogaraphs of used in the previous two subjective evaluations, ex-
actual meter clusters cluding expressions that were thought to be difficult
Subjects were asked to evaluate photographs of to evaluate, and words that the design department
actual meter clusters for the purpose of analyzing the asked to have included. The following is a list of the

Table 3
Results of analysis ('luxurious', 'clean looking')

Design Partial Not luxurious Luxurious Partial I Cluttered Clean-looking

Category [correlation Partialregression correlanon[ Partialregression
element coefficienJt " ~ e f f i c i e n l 0.5 coetII
coettmtentl coefficient 0.5
l. L~=-J i -0.243 ~-0.794~
Scale
2. ~ [ 0.49 0.124 0.91 I -0.353
3.'' 'i -0..72 m0.93 l ~ m
_ _ ~ .......... i i 0.035 ~-1.519i
1. 0123 !-0.5671 -0.005
Letterin~ 3.2" 0Of.go.1
2 3 I 0.92 0.166 0.39 0.188
~"~'0.8001 -0.248
4. 0 1 2 3 -0.066 m //.065
-0.044 I R 0.092
2. ~ I 0.63 m 0.098 /I.067
-0.157~ -0.043
m-0.489~ 0.518 I , I
0.100 0.198
0.053 0.52 4).276
-0.247 -0.318
 T. Jindo, K. Hirasago / lnternational Journal of Industrial Ergonomics 19 (1997) 105-114 109

Table 4
Design elements and categories of each sample evaluated

sign Overall cluster layout Meter shape Indicator color

(no. of meters) Scale

3 4 5 or more
meters meters meters OC) I ] 1, l , I h,,,u,,I White Cellow Red

1. 0 0 0 0
2. 0 0 0 ©
s. 0 0 0 0
4. 0 0 0 (D
5. 0 0 0 ©
6. 0 0 0 0
7. 0 0 L--
0 0
8. 0 0 , 0 0
9. 0 0 _~ 0 0
10. 0 0 0 0

Table 5
Quantification I method analysis results for 'sporty'

Partial Partial
Design Category correlation regression Not sporty -. ~- Sporty
element
coefficient coefficient -o;5 (1.5
I

1. Overall 1.3 meters 0,79 -0.432

cluster layout 2.4 meters -0.749
(no. of meters 3.5 or more
meters
2. Meter shape l. R o u n ~ 0.49 i
2. Semioir~_ -0.139 u
cular ~_~
3. Scale 1.1 1 I 0.11 -0.181 l
2.1~Itl @ m
3. I~,1,,,[ 0.155 m
4. Indicator 1. White ] 0.46 -0.263 n
color 2. Yellow m
3. Red 0.052 in
110 ~ Jindo, K. Hirasago/ International Journal of Industrial Ergonomics 19 (1997) 105-114

~-to-understand axis1
• easy-to-understand

• clean-lookang

@
• Retro-looking eElegant

•LikabIe Fig. 6. A sample of meter cluster.

•Luxurious

• Sporty layout having five or more meters, a round meter

• Playful
shape, a scale with medium-level graduations and
yellow-colored indicators.
Fig. 5. Arrangement of evaluation adjectives and meter cluster
samples. 2.1.4. Conclusions q f speedometer study
In addition to the subjective evaluations and anal-
expressions used: easy-to-understand, practical, luxu- yses described above, other evaluations were also
rious, sporty, subdued, enthusiast-oriented, appeal- conducted under different test conditions using vari-
ing, clean-looking, elegant, well-laid-out, well- able design elements. The results of those evalua-
shaped, well-organized and likable. A total of 30 tions are also reflected in the following conclusions.
subjects participated, divided equally between men I. Two independent factors that appear to influence
and women. the static impression conveyed by automotive in-
The results obtained for the ten samples are sum- strumentation are the design and a feeling of
marized in Table 4 in relation to the design elements being easy to understand. In nearly every case,
and their categories. Similar to the procedure em- the results of the series of subjective evaluations
ployed for the subjective evaluations of speedome- conducted in the present work could be inter-
ters, the Quantification I method was used to analyze preted in terms of these two factors.
the relationships between the evaluation scores and 2. Correlations were made between the design ele-
the design element categories in Table 4. Typical ments of analog meter clusters and the evaluation
analytical results obtained for the adjective 'sporty' adjectives. It was confirmed that the correlations
are shown in Table 5. The results in Table 5 indicate could be used in determining meter cluster speci-
that a sporty impression was conveyed by an overall fications that would convey an intended image.

Table 6
Classification and definition of design elements and categories
Design elements Category Definition
The numberof spokes 1.2 The numberof spokes is 2.
2.3 The numberof spokes is 3.
3.4 The numberof spokes is 4.
Pad surface shape 1. Flat Pad surface is flat.
2. The secondorder curved surface Pad surface is of the second order curved surface.
3. The third order curved surface Pad surface is of the third order curved surface.
Area of pad 1. Large Pad area is large.
2. Medium Pad area is medium.
3. Small Pad area is small.
Pad upper side shape 1. Sharp projectionshapeR in upper side is small.
2. Gentle projectionshape R in upper side is large.
3. Varied projection shape R in upper side is varied.
 T. Jindo, K. Hirasago / International Journal of Industrial Ergonomics 19 (1997) 105- 114 111

2.2. A study o f steering wheel on Kansei engineering cation 1 method. This analyzing method is used for
categorical data including external criteria. In this
2.2.1. Subjective evaluation on steering wheel case, the dependent variables are the ratings for 41
After the research on speedometers we investi- pairs of evaluation terms as external criteria and the
gated steering wheels. descriptive variables are the design categories as
We carried out subjective evaluation tests on categorical data.
steering wheels to relate their impressions to physi- Fig. 7 shows the result of the analysis of the
cal features. In these tests, samples of 59 types of impression evaluation concerning 'sporty' or 'not
steering wheels, shown by projector films, are ranked sporty'. The multiple correlation coefficient in this
by 50 persons according to SD 5-grade method figure indicates to which degree a certain item and
regarding 41 pairs of evaluation terms. its category classification which are chosen as design
We prepared those test samples by putting each elements can explain the evaluation result of the
photograph of automobile steering wheels on the relevant term. The partial correlation coefficient
market into Macintosh computers using a scanner, shows the correlation between each item and the
and then masking the backgrounds and eliminating evaluation, and the category score shows the one
corporate marks on the steering wheels by graphic between each category and the evaluation. To give
processing of 'Photo Shop', an image processing an example from Fig. 7, a sporty image is empha-
software. sized by the following design items: three or four
Prior to the analysis of the test data, we classified spokes, the second order curved surface of the pad
the steering wheel designs to investigate how the and a smaller pad area.
ratings of the evaluation terms are influenced by Conducting this analysis for every evaluation term,
such design items as the number of spokes and pad relating the results of the evaluation to each physical
size. In this classification, we chose 13 design items feature (design), and storing these results in a
which are supposed to influence human impression, database will allow us to acquire the knowledge to
and established 34 categories consisting of two or incorporate human sensitivity into products.
three categories for each design item. Table 6 shows
a part of this item category classification. 2.2.2. Construction o f image retrieval system
We analyzed the correlation level between the This section describes the image retrieval system
rating results obtained from the tests and the item for steering wheels we have constructed into which
category classification shown in Table 6 by Quantifi- analysis data of the test results are incorporated as a

Multiple correlation coefficient=0.7605

Partial Not sporty -~ ~ Sporty
Design Category correlation
.1.o Category score 1.o
Elements coefficient

The nurnbe 1.2 -1.296

2.3
o f spoe,es 3.4 0.728
/ 0.410
1. Flat -0.146 •
Pad 2. The second order
surface curved surface 0.693 0.694
shape 3. Thethirdorder -0.518 1
curved surface
1. Large -0.212 1
Area of 2. Medium 0.618 -0.265 1
pad 3. Small
1.Sharp projection shape
Pad upper 2. Gentle projection shape -0.369 1 0.685
side shape 3. Varied projection shape -0.242 1

Fig. 7. Results of quantitative analysis.

112 T. Jindo, K. Hirasago / lnternational Journal of Industrial Ergonomics 19 (1997) 105-114

In'P-put
of adjective

dImage
--ata processing ~ - - ~

Pad
surface ]r"eseondorder
curvedsurface 2 t
Inference of priority o f ~ . _ ~ shape 15. ]:hesecondorder . . . .
sample candidates ~ ~ ---4 curvedsurface _ Q) i)
areao~ I!:L~e ~ -- I I / c'
pad ~ 2 . Medium O
l•Graphic display
~ of s a m p l e s ~ - - ~

Fig. 9. Sample design database.

Fig. 8. System flowchart.

database. This system, constructed on Macintosh

computers using THINK C, connects the evaluation item classification table regarding the evaluated
terms used in the tests with the steering wheel steering wheels. Fig. 9 shows a part of the table.
images, making it possible to retrieve and check The sample PICT database is a PICT-type storage
concrete samples in response to inputted evaluation of the image files of steering wheels used for the
terms. Fig. 8 displays the system flow chart. tests. The processing in the system is conducted as
The image database includes a partial correlation follows. When an adjective is entered, the data (par-
coefficient and category score (partial regression co- tial correlation coefficient and category score) for
efficient) of design items in each evaluation term such adjectives is read from each image database.
which are obtained from the analysis of the test Each sample's item category information is read
result. There are three types of database because the from the sample design database. (See Fig. 9.) Image
data analysis was carried out for three subject groups scores of each sample (the sum of each item's partial
consisting of male, female and combined. correlation coefficient multiplied by each item's cat-
The sample design database is built as a design egory score) are found. Then, calculated image scores

Fig. 10. Output screen of 'serene' steering wheel.

 T. Jindo, K. Hirasago / lnternational Journal of lndustrial Ergonomics 19 (1997) 105-114 113

of each sample are arranged in descending order to tion adjective and decide on unit styling from the
be regarded as the candidate priority. After reading unit impressions database when an operator inputs
the steering wheel PICT images according to this total evaluation adjectives.
candidate priority, the graphic display is started. Fig.
lO shows an example of the output when 'serene' is
inputted. 4. For further reading

2.2.3. Conclusions of steering wheel study Ishihara et al., 1993; Jindo et al., 1990; Jindo and
In this study, subjective evaluations about the Nagamachi, 1991; Jindo et al., 1994a,b,1995; Naga-
relation of the impressions of steering wheels to their machi, 1989, 1986; Shimizu et al., 1989; Yanag-
features were carried out. Then we analyzed the ishima and Nagamachi, 1988.
results of the evaluation, quantitized that relation,
and finally constructed an image retrieval system
which output the best fit photograph on the CRT Acknowledgements
according to the adjective input. In other words, we
proposed an environment where the persons from the This study was conducted collaboratively by Nis-
design department and styling department are able to san Motor Co., Ltd. and Mr. Nagamachi's office of
discuss the evaluation of the steering wheel with Hiroshima University. Students of Hiroshima Uni-
each other using common adjectives. versity helped us in evaluating the film samples in
this study. We wish to express our gratitude to all of
the persons concerned.
3. Conclusions

In order to get more detailed impression data

References
indicating the relation of evaluation scores to its
styling features, experiments and analyses of small lshihara, S., Hatamoto, K., Nagamachi, M. and Matsubara, Y.,
units of passenger car interiors rather than whole 1993. An automatic builder for kansei engineering expert
units were carried out. As a result, the detailed system using self organizing neural networks. In: C. Gu and
impression data also seemed to be useful for actual H. Osaki (Eds.), Proc. Second China-Japan International Sym-
styling. posium on Industrial Management. International Academic
Publishers, Beijing, pp. 497-502.
Impression studies of interior units will proceed in Jindo, T., Yanagishima, T. and Shimizu, Y., 1990. An evaluation
the future. Accumulating impression data of impor- structure of automobile interiors using multivariate analysis.
tant units constituting car interiors will be achieved. 23rd Fisita Congress Technical Paper, pp. 635-641.
However, there is the problem where an evalua- Jindo, T. and Nagamachi, M., 1991. The development of car
tion adjective of a certain unit is not always used in interior image system incorporating knowledge engineering
and computer graphics. In: Y. Queinnec and F. Daniellou
the same meaning or nuance as in a whole interior (Eds.), Proc. llth Congress of the International Ergonomics
evaluation. Association, pp. 625-627.
For example, the car interior that is evaluated Jindo, T. et al., 1994a. A study of kansei engineering method
as'comfortable' must have an 'easy-to-understand' applying to car interior. Human Factors in Organizational and
speedometer, ' not oppressed' dash shape, and 'ex- Management, IV: 479-484.
Jindo, T. et al., 1994b. A study of kansei engineering on steering
cellent' seating. wheel of passenger cars. In: Proc. 1994 Japan-U.S.A. Sympo-
Exactly, evaluation adjectives of each unit may sium on Flexible Automation, Vol. 2, pp. 545-548.
not have the same meaning if the same evaluation Jindo, T., Hirasago, K. and Nagamachi, M., 1995. Development
adjective is used in whole interior evaluations. of a design support system for office chairs using 3-D graph-
The future task involves examining relations of ics. International Journal of Industrial Ergonomics, 15(1):
49-62.
total evaluation and unit evaluation adjectives. Nagamachi M., 1989. Sensory engineering approach to automo-
We will reconstruct styling support systems aimed bile. Journal of the Society Automotive Engineers of Japan,
at whole car interiors which select each unit evalua- 43(1): 94-100.
114 T. Jindo, K. Hirasago / lnternational Journal of Industrial Ergonomics 19 (1997) 105-114

Nagamachi, M., 1986. Emotion technology and its application. Yanagishima, T. and Nagamachi, M., 1988. Sensory engineering
The Japanese Journal of Ergonomics, 22(6): 319-324. approach to automobile interior. The Japanese Journal of
Shimizu, Y., Nagamachi, M. and Jindo, T., 1989. Analyses of Ergonomics, 24(6), 38-40.
automobile interiors using a semantic differential method. In:
A.D. Keller (Ed.), Proc. Human Factors Society 33rd Annual
Meeting, Vol. 1, pp. 620- 624.