Colour Emotions for Real Objects.

Colour patches have been widely used by designers in developing colour schemes and sometimes by customers in colour selection. It is important to clarify whether colour patches show the same emotional effect as those by real objects. In our previous study, several colour science based colour-emotion models were developed for colour patches. To see whether these models can be applied to real objects, a psychophysical experiment was carried out in a real-sized room where the following four items were presented: coloured vases, coloured walls, tablecloth, and carpet. The experiment was divided into ten sessions and the walls were painted ten times with different colours. In each session, ten vases in different colours were presented individually in front of the walls. Eleven adjective pairs were used in the experiment for subjects to judge colour emotions evoked by the combination of the vase and the wall colours. The experimental results show little impact of the tablecloth and the carpet colours on the colour emotions for the whole scene of the room. The results also show that in this real-room environment, the foreground colours (the vases) made far greater contribution to colour emotions of the whole scene than did the background (the walls). It should be noted that the wall colours can be regarded as a surround rather than a background, and thus it is difficult to tell whether the experimental results were owing to the fact that there was no “foregroundbackground effect”. Therefore, it would be necessary to clarify the difference between the effects of background and surround for the future work.

Li-Chen Ou Ming Ronnier Luo University of Leeds Andrée Woodcock Coventry University

174

Colour Emotions for Real Objects Colour patches have been widely used by designers in performing colour planning and sometimes by customers in making colour selection. It is thus important to know whether a colour patch have the same emotional effect as those by a real object. In our previous study, colour science based colour-emotion models were developed for colour patches. To see whether these models can be applied to real objects, a psychophysical experiment was carried out in a real room where the four items were presented: coloured walls, coloured vases, tablecloth, and carpet. The experiment was divided into 10 sessions and the walls were painted 10 times with different colours. Ten vases with different colours were presented one at a time in front of the walls. Eleven adjective pairs were used for subjects to assess colour emotions evoked by the combination of vase and wall colours. Experimental results show that the tablecloth and the carpet colours had little impact on colour emotions for the whole scene. The results also show that in a real-object environment, a foreground colour makes far greater contribution to colour emotions of the whole scene than does a background colour. Nevertheless, since the wall colours can be regarded as a surround, it is difficult to tell whether the experimental results were owing to that there was no “background effect”. It is thus necessary to clarify the difference between background and surround effects in future work. Key words: colour emotion, colour meaning, colour planning, colour design, real-object environment, foreground-background effect

INTRODUCTION Colour patches have been widely adopted by researchers in studying colour preference and colour emotion, and also by designers in performing colour planning for a product. Sometimes they are used by customers in making colour selection. The question is whether the colour emotions evoked by a colour patch have the same emotional impact as those by a real object. In comparisons between colour emotions for colour patches and for real-object images, Taft (1997) demonstrated good correspondence between the two display media. The comparisons were made on a Cathode Ray Tube (CRT) display, where both colour patches and real-object images were presented. Note this study did not make comparison between colour patches and real objects, but between colour patches and simulated real-object images. Hoshino et al. (1997) made comparisons of colour emotions between colour patches (physical samples), real room doors, and digital images of the room doors. The results showed little difference between the three display media. The studies above all focused on single-colour emotions. In a real-object environment, however, colours never exist in isolation. A real-object colour may 1

be affected by contiguous colours and also by background colours in terms of emotional impact. Helson and Lanford (1970) in their colour pleasantness study indicated that background colours exerted a significant effect on pleasantness of foreground-background colour combinations. Camgöz et al. (2002) showed that the three colour-appearance attributes, hue, lightness, and chroma, all had influence on colour preference for foreground-background combinations presented on a CRT display. Note that these two studies were both using colour patches as stimuli. In the previous studies (Ou et al. 2004a-b) we developed colour science based models to predict a colour-emotion value of a colour pair by the mean value of the component colours in that pair, where the single-colour emotions can be predicted from the three colour-appearance attributes, hue, lightness, and chroma. Note that the models developed were all based on colour patches, and accordingly they do not necessarily apply to viewing conditions that are deviated from these studies. The present article investigates whether colour emotions of real objects can be predicted by colour patches. The foreground-background effect on colour emotions is also discussed. METHODS To investigate colour emotions for real objects, a psychophysical experiment was carried out in a real room, in size of 4 m (length) by 3 m (width) by 3 m (height). As shown in Fig 1, ten coloured vases were presented one at a time on a fabric-covered table at a corner of the room. The room floor was covered with a dark-grey carpet. This room was illuminated by a D65 daylight simulator situated under the ceiling, which was painted white. In the experiment the room door was kept closed and the window was covered with a black blind in order to stop outside light from coming into the room. The vases were painted with the 10 colours: red (labelled V1), orange (V2), yellow (V3), blue (V4), green (V5), bluish green (V6), pink (V7), brown (V8), black (V9), and grey (V10). These colours were selected to give a reasonable coverage of colour range in terms of hue, lightness, and chroma, as shown in Fig 2 (a). The 4 walls of the room were all painted with a single colour for each experimental session. There were 10 sessions in the experiment, and thus 10 wall colours were used, labelled W1 to W10, as shown in Fig 2 (b). Each of the vases was presented on a table covered with a piece of light-grey tablecloth. Under the table was a dark-grey carpet that covered the entire floor. Both the tablecloth and the carpet were kept constant for all experimental sessions.

2

Colorimetric values of the vases and the walls were summarised in Tables 1 (a) and (b), respectively. These values were determined by measurement of a Minolta CS1000 tele-spectroradiometer, at the position of subjects’ eyes, a height of 115 cm and a distance of 250 cm from the vase. The measurement was carried out before each experimental session. Coloured Carpet

250 cm Subject

FIG 1. Experimental room: top view b*

b* 120

120 W10 V3

80

W9

80 V2 40

V5

40

V8 V10 0

V6

V9

W2

V1 0

a*

V7

V4

-40

-80 -100

-50

W8 W6

50

-80 -100

100

(a)

W5

a*

W7 W4

-40

0

W3

W1

-50

0

50

100

(b)

FIG 2. The colour stimuli in the CIELAB colour space (Hunt, 1998) for (a) vase colours (V1 to V10) and (b) wall colours (W1 to W10). The axes a* and b* represent reddish-greenish and yellowish-bluish, respectively.

Table 1. CIELAB values of (a) vase colours and (b) wall colours. L*, C*, and h stand for the three colour appearance attributes, lightness, chroma, and hue angle, respectively. (a) Vase

description

V1 V2 V3 V4

red orange yellow blue

3

L*

C*

h

38.3 57.0 71.2 42.3

51.0 60.9 76.7 35.1

23 66 85 248

V5 V6 V7 V8 V9 V10

green bluish green pink brown black grey

44.7 35.3 69.6 38.0 22.2 69.3

33.6 22.4 10.6 17.4 0.6 1.2

117 217 316 52 296 10

Wall

description

L*

C*

h

W1 W2 W3 W4 W5 W6 W7 W8 W9 W10

white pale yellow pink pale blue more colourful pink pale cyan pale purple pale green bright orange bright yellowish green

88.1 84.7 79.7 66.9 71.5 64.4 81.0 73.4 87.1 80.3

7.4 22.5 12.3 24.0 25.1 17.4 9.6 30.0 29.1 24.7

100 92 24 254 21 208 296 166 84 105

(b)

Eleven adjective pairs, also called colour-emotion scales, were used in the experiment, including warm-cool, heavy-light, modern-classical, dirty-clean, active-passive, hard-soft, tense-relaxed, fresh-stale, feminine-masculine, like-dislike, and harmonious-disharmonious. These adjective pairs were exactly the same as those used in our previous study (Ou et al., 2004a). The experiment was divided into 10 sessions. Before each session 2 to 3 weeks were taken to paint the room and to wait for the paint dry and the odour disappeared. The entire experiment took a year to complete, from Dec 2001 to Jan 2003. A total of 20 subjects were involved in the experiment, while only 10 of them attended more than half of the sessions. In each session 9 to 11 subjects undertook the experiment. The subjects were either staff members or students of the Colour & Imaging Institute, University of Derby (UK), including 5 British and 15 Chinese. They all passed Ishihara’s Tests for Colour Deficiency (Ishihara, 2003). The method of categorical judgement was used for data collection. Each subject was given the question, “Considering the combination of the wall and vase colours, which of the following words associates with the combination—warm or cool?” For each vase the 11 adjective pairs were judged in random order. In each judgement, say warm-cool, subjects were asked first to give an answer of either warm or cool and then to indicate how warm/how cool it appeared by giving a number from 1 to 5, where 1 the weakest and 5 the strongest. Note that subjects were not given the option of “uncertain” as a response. 4

All the questions and answers were given orally between the experimenter and the subject. One of the two languages, English and Chinese, was used in the experiment according to subject’s native language. The subjects were shown definitions of the 11 adjective pairs before experimental sessions. The definitions were made in reference to Cambridge Advanced Learner’s Dictionary. RESULTS AND DISCUSSION According to Torgerson’s Law of Categorical Judgement (Torgerson, 1958), the experimental data were transformed into categorical scale values, called colour-emotion scores in the present study. A colour-emotion score indicates the extent to which an adjective pair agreed with a vase-wall combination. The Background Effect The colour-emotion scores were compared between experimental sessions to see how different wall colours, regarded as the background, affected the visual data (the colour-emotion scores). As shown in Table 2, the comparisons were made on the 11 colour-emotion scales, where all the scales show fairly high correlation of visual data between experimental sessions, except for “harmonious-disharmonious” and “like-dislike”, with mean Pearson correlation values of 0.08 and 0.35, respectively. This indicates that in the 2 scales the wall colours significantly affected the visual results, while the effect was little in the other scales. Table 3 also shows little difference of colour-emotion scores between experimental sessions. Table 2. Mean values of Pearson Correlation of visual data between experimental sessions on each colour-emotion scale Warm-

Heavy-

Modern-

Dirty-

Active-

Hard-

Harmonious-

Tense-

Fresh-

Feminine-

Like-

cool

light

classical

clean

passive

soft

disharmonious

relaxed

stale

masculine

dislike

0.97

0.94

0.79

0.88

0.89

0.84

0.08

0.85

0.93

0.77

0.35

Table 3. Pearson Correlation of visual data between experimental sessions (W1 to W10) W1 W2 W3 W4 W5 W6 W7 W8

W1 -

W2 0.82 -

W3 0.77 0.86 -

W4 0.78 0.80 0.78 -

W5 0.78 0.80 0.86 0.82 -

W6 0.82 0.81 0.81 0.85 0.73 -

5

W7 0.82 0.79 0.83 0.87 0.83 0.83 -

W8 0.81 0.82 0.83 0.86 0.79 0.88 0.84 -

W9 0.85 0.86 0.86 0.76 0.85 0.81 0.80 0.82

W10 0.76 0.85 0.88 0.81 0.79 0.83 0.82 0.87

Mean 0.80 0.82 0.83 0.81 0.81 0.82 0.83 0.84

W9 W10

-

0.86 -

0.83 0.83

Testing Colour-emotion Models In the previous study (Ou et al., 2004a) 4 colour-emotion models were developed for colour patches, warm-cool, heavy-light, active-passive, and hard-soft, as given in Eqs (1) to (4), respectively. These models predict colour-emotion values for single colours if the three colour-appearance attributes, hue, lightness, and chroma, are known.

WC = −0.5 + 0.02(C*)1.07 cos(h − 50°)

(1)

HL = −2.1 + 0.05(100 − L*)

(2) 1

⎡ ⎛ L * −50 ⎞ 2 ⎤ 2 2 AP = −1.1+ 0.03⎢(C *) + ⎜ ⎟⎥ ⎝ 1.5 ⎠ ⎦ ⎣

(3)

HS = 11.1 + 0.03(100 − L*) − 11.4(C*) 0.02

(4)

where L* is CIELAB lightness; C* is CIELAB chroma; h is CIELAB hue angle. In another study (Ou et al., 2004b), a model was developed to predict a colour-emotion value for a colour pair by averaging colour-emotion values of component colours in that pair. For a combination of more than 2 colours, we assume the colour-emotion value of the whole combination be determined by E=

1 ( E1 + E 2 + L + E N ) N

(5)

where E is the colour-emotion value of a colour combination; E1, E 2 , L E N are colour-emotion values of component colours in that combination; N is the number of component colours. Eq (5) states that all component colours have equal contribution to the whole combination, provided that all the colours have the same shape, texture, size of area. In the present experiment the contribution of each colour stimulus, i.e. the vase, the walls, the tablecloth, and the carpet, were obviously different from each other. Thus the colour emotion of the whole scene can be described by

E = fv Ev + fw Ew + ft Et + fc Ec

(6)

6

where E is the colour-emotion score obtained from the experimental data; E v , E w , E t , E c are colour-emotion values of vase, walls, tablecloth, and carpet, respectively, which can be predicted by existing models such as Eqs (1) to (4); f v , f w , f t , f c are constants and f v + f w + f t + f c = 1. The 4 coefficients f v , f w , f t and f c were determined by minimising the difference between visual data and predictive values. This was performed by the use of the Microsoft Excel Solver Tool, which suggests 0.77 for f v , 0.23 for f w , and 0 for both f t and f c , as summarised in Eq (7).

E = 0.77E v + 0.23E w

(7)

The 4 coefficients can be regarded as having a ratio of 3:1:0:0, suggesting that in the prediction of colour-emotion scores both the tablecloth and carpet colours can be neglected and the contribution of vase and wall colours was 3 to 1. Nevertheless, Tables 1 and 2 both show that the wall colours had little influence on the colour-emotion scores. This indicates that although the optimisation result shown in Eq (7) indicates a ratio of 3:1, the equation does not perform significantly better than the prediction made only by using colour-emotion value of the vase ( E v ). This also suggest that in a real-object environment the foreground colour tends to have predominant impact on colour emotion of the whole scene, even if the foreground occupies a smaller area than the background. CONCLUSIONS Experimental results show that all the colour-emotion scales, except “harmonious-disharmonious” and “like-dislike”, can be predicted for vase-wall combinations by the models for colour patches, suggesting that colour emotions were not influenced by the use of different display media. This agreed with the early studies (Taft, 1997; Hoshino et al., 1997), which are mentioned earlier in the present article. Experimental results show that the two scales “harmonious-disharmonious” and “like-dislike” were affected in visual data by the use of different wall colours. This again agreed with the findings by Helson and Lanford (1970) and Camgöz et al. (2002) in which they found background colours exerted a significant effect on pleasantness and preference of foreground-background colour combinations. The present study finds that the tablecloth and carpet colours can be neglected in predicting colour emotions for the whole scene of the room. This was perhaps because in the experiment the subjects unconsciously focused on vase-wall colour combinations. Experimental results suggest that in a

7

real-object environment, a foreground colour makes far greater contribution to colour emotions of the whole scene than does a background colour. This is because the foreground draws almost all the attention of the viewer. Note, however, that the wall colours were not only a background, but also a surround. Fig 3 illustrates the relationship between foreground, background and surround. Note that the term “surround” here is used on a common-use basis and is not defined as that in the field of Colour Appearance (Hunt, 1998). In the present study, accordingly, it would be better to say that the experimental room was a foreground-surround environment rather than a foreground-background one. It is thus difficult to tell whether the consistent colour-emotion scores across different wall colours were resulted from that there was no “background effect”. Therefore, it is important for future studies to clarify the difference between “background effect” and “surround effect”. Surround Background

Surround

Foreground

Surround

Surround

Fig 3.

The relationship between foreground, background, and surround

REFERENCES Camgöz N, Yener C, Güvenç D. (2002) Effects of hue, saturation, and brightness on preference. Color Research and Application 27: 199-207. Helson H, Lansford T. (1970) The role of spectral energy of source and background color in the pleasantness of object colors. Applied Optics 9: 1513-1562. Hoshino H, Nakamura T, Sato T, Kajiwara K, Miki Y, Shino T, Kuwano K. (1997) Color planning at school through visual assessment. Proceedings of AIC Color Congress 1997: 972-975. Hunt RWG. (1998) Measuring Colour. Third Edition. Hertfordshire: Fountain Press.

8

Ishihara S. (2003) Ishihara’s Tests for Colour Deficiency. 38 Plates Edition. Tokyo: Kanehara Trading Inc. Ou L, Luo MR, Woodcock A, Wright A. (2004a) A study of colour emotion and colour preference, Part I: colour emotions for single colours. Color Research and Application 29: 232-240. Ou L, Luo MR, Woodcock A, Wright A. (2004b) A study of colour emotion and colour preference, Part II: colour emotions for two-colour combinations. Color Research and Application, in press. Taft C. (1997) Color meaning and context: comparisons of sementic ratings of colors on samples and objects. Color Research and Application 22:40-50. Torgerson WS. (1958) Theory and Methods of Scaling. John Wiley & Sons.

9