Emotion-Spectrum Response to Form and Color: Implications for Usability Alan Manning Brigham Young University [email protected]
Abstract Previous empirical studies have shown consistent emotional responses to form and color, across a variety of contexts and especially across cultures. What varies from context to context and culture to culture is the evaluation of the color/form/emotion response. For example, both the color red and jagged, high contrast forms consistently evoke one emotional response neutrally described as agitation or activation, a response evaluated negatively as anger or positively as excitement. Standard taxonomies of emotion do not consistently distinguish between the evaluation of an emotion and its raw quality. Consequently, the consistent relationships between form/color and emotion have been obscured; evaluative emotional descriptions of color/form do vary from person to person, context to context, and culture to culture. We propose a new model of emotional response which treats color/form triggers of emotion quality separately from triggers of emotion evaluation. This new model identifies a spectrum of emotional quality (agitated-stimulated-diverted-calmed-organized-focusedconcerned) generally parallel to the familiar color spectrum (red-orange-yellow-green-blue-indigo-violet). With this model, we can demonstrate a stable emotionspectrum response in a population of viewers, to any given combination of form and color. This paper will report on empirical tests of this emotion spectrum model and discuss implications for usability testing of visual information designs.
Introduction: emotion and evaluation Citing black as both a color of mourning in the U.S. versus the bride's wedding color in parts of Europe, Kress and Van Leeuwen  pessimistically evaluate semantic models of color, assuming that ideational meaning is the kind of meaning that matters when talking about color (red=stop, green = go, white=good, black=bad, etc.). Ideational color meaning varies from context to context, but that is equally true of ideational word meaning (i.e. cup=drinking vessel, cup=sports equipment, etc.). In contrast to the ideational meaning of color, this study will examine the affective aspect of color meaning, in other words the raw emotional response. There have been a number of empirical studies (e.g. by Collier , Zetner , Ou et. al. , and Xin et. al. ) showing consistent emotional responses to form and color, across a variety of contexts and especially across
1-4244-1243-9/07/$25.00 ©2007 IEEE
Nicole Amare University of South Alabama [email protected]
cultures. In these studies, what varies most about color meaning is the evaluation of the color/form/emotion response, as observed by Oyama: Cross-cultural comparisons indicated wide generality of affective and symbolic meanings of colors and forms. However, some cultural differences were found, especially in evaluation. Affective meanings of various color combinations and colored forms were compared with those of component colors and forms ... but interaction between components were found important especially in the evaluative dimension. [6, p. 137] In short, the emotional response to, for example, a composite figure that combined red and yellow was found to correlate with the emotional response to both red alone and yellow alone (as high as .96 for the "active" emotional interpretation of these colors), but this correlation is partly obscured by positive or negative evaluation (reducing the correlation to .55), whether the subjects happened to evaluate the exact form of the combination as good or bad, ugly or attractive [5, p. 138]. Such results point to a significant widespread problem in color-meaning research: the model of emotional responses to color has been constructed primarily of terms with a strong evaluative component. That is, terms like angry, sad, or ugly imply a negative evaluation, while pleased, happy, or attractive imply a positive evaluation. In addition many such "emotion" terms have also involved an ideational component (i.e. a necessary mental picture in addition to a raw feeling) with can further obscure consistent color-emotion responses, as for example, in the emotion-term inventory offered to test subjects by Ou et. al. [4, p. 236] – Four word-pairs that relate more or less directly to raw feeling or raw emotion: Warm–cool, Heavy–light, Active–passive, Hard–soft, Two word-pairs that involve ideational concepts as well a feeling: Modern–classical, Masculine–feminine, And four word pairs that involve both evaluation (a sense of good vs. bad) and ideational concepts:
Clean–dirty, Like-dislike, Tense–relaxed, Fresh–stale. Ou et. al. [4, p. 238-239] only report reliable correlations of color and emotion for the terms that we here identify as being free of both ideation and evaluation: Warm–cool, Heavy–light, Active–passive, and Hard–soft. This result compares with those of Oyama et. al. cited above. The studies cited so far (and a good many others) have all relied on a single theoretical model of emotion proposed back in 1957 by Osgood et. al. . Collier's 1996 study found the Osgood model only suitable as a "first approximation" for mapping color and form to emotional response, noting that further refinements to the theory of emotion were needed to account for his experimental results [2, p. 1]. It is also difficult to derive practical informationdesign guidelines from the various empirical studies based on the Osgood et. al. model of emotion. Practically minded designers of advertisements, web pages or software interfaces seem unlikely to wish to communicate "warmth" per se, or "coolness" or "heaviness" or "softness," these being tactile metaphors for common emotions rather than precise descriptions of the emotions themselves. It seems more precise to say that designers might wish to make viewers feel the urge to act, i.e. feel agitated or stimulated. They might instead wish viewers to feel a sense of fun or freedom, i.e. feel diverted or amused. They might instead wish to make viewers feel settled or at peace, calmed or organized, focused on, concerned with, or challenged by some issue. It would therefore be useful to know what strategies of color choice and colorform combinations can reliably produce these emotional effects, even in a single culture if not cross-culturally. However, no theory proposed thus far has this kind of practical reach, as noted by Puhalla (2008): Color is used as a coded information system in design, architecture, cartography, science, medicine, industry and government. However, principles for the application of color in these systems have little documented research as their basis. Due to the lack of scientific research in the area of color organization, objective color structuring has a weak theoretical base and no means of producing predictive value. Simple laws of color organization are believed to be skewed by human subjective responsiveness. [8, p. 200] In this paper we will propose a refined theoretical model of emotion, capturing the range of practical emotions described above with dimensions based on the primary categories from the model of meaning developed by C.S. Peirce . We will review results from our initial empirical test of this model, evaluating it in terms of its ability to capture reliable correlations between color, form, and emotion. We will also discuss direct
implications of our results for information design and usability testing.
Method: theory and experiment Peirce's central claim was that all of our perceptual, physical, and intellectual experiences decompose into simple or complex configurations of three primary categories, which he rather abstractly designated as firstness, secondness, and thirdness. For purposes of discussion here, we will more concretely reference these same categories as (1) potential, (2) reaction, and (3) pattern. In theory then, potential, reaction, and pattern are the building blocks of all experiences, just as three primary colors can be combined in different ways to create different shades of color, ultimately millions of distinctions detectable to the human eye. There are likewise indefinitely many shades of possible emotion or feeling. Just as millions of colors are still classifiable in broad ranges of primary colors and their secondary combinations (red-orange-yellow-greenblue-purple), all these emotional shades would be classifiable within the Peircean framework in the range of the primaries or secondary combinations of those primaries: emotions felt as potential, felt as reaction, and/or felt as pattern. Anger for example we would construe primarily as reaction (Peircean category 2, evaluated negatively). Obsession for example we would construe as a combination of reaction and pattern (Peircean categories 2+3, evaluated negatively). That is, when we feel obsessed, we feel obsessed about (i.e. in reaction to) something, a repetitive reaction that will persist (i.e. as a pattern). Evaluated more neutrally, if someone primarily reacts to something, we would say that person feels agitated or challenged. If someone reacts in a patterned, perpetual way, we would say that person is focused on or concerned about that thing. The rest of our postulated range of emotions, with their Peircean category analysis, is summarized in Table 1. Table 1. Range of Emotions as Peircean Categories Neutral
diverted stimulated agitated focused organized calmed
entertained aroused active committed reliable peaceful
distracted anxious angry obsessed trapped bored
1=potential 1+2 2=reaction 2+3 3=pattern 1+3
yellow orange red purple blue green
Peirce's first category, potential, defines the neutral, positive, and negative values of category-1 emotion in Table 1. That is, to feel diverted is to feel a range of possibilities or potentials, not feeling focused on or driven to one course of experience or action. To be entertained is to be diverted by a range of potentials in a
"happy" way. To be distracted is to be diverted by a range of potentials in an "unhappy" way. Note that category 1 emotion (diverted-entertaineddistracted) is the structural complement and semantic opposite of the category 2+3 emotion (focusedcommitted-obsessed), in exactly the same way that yellow is the color complement of purple, but this is perhaps most apparent in the semantically neutral descriptions, diverted vs. focused. Complement/opposite relations also hold that are analogous to red/green and orange/blue, between agitated (2) vs. calmed (1+3), and between organized (3) vs. stimulated (1+2) in the sense that stimulation always disrupts existing organizations and patterns in some way. In theory, all other conceivable emotions can be placed somewhere with in the ranges indicated in Table 1. Positive and negative emotion values might be considered roughly equivalent to light and dark values of color. High and low intensities of emotion might be considered roughly equivalent to high and low saturation levels of color. However, our main concern in this study is the relationship between the postulated emotional spectrum, agitated-stimulated-diverted-calmed-organized-focused, and the familiar color (i.e. temperature) spectrum, redorange-yellow-green-blue-purple. Our first hypothesis is that these neutrally-described emotions correlate reliably their model colors, especially in the absence of distracting factors such as positive/negative evaluation and ideational association, such as the connection between the color blue and water, between the color red and communism, etc. Our second hypothesis is that these neutrally described emotions also correlate with distinct form qualities, formal variety correlating with diversion-range emotions, formal contrast correlating with agitation-range emotions, and formal repetition correlating with organization-range emotions. In practice, it is impossible to separate color from form, since even a large region of solid color has a form (with low contrast and possible negative correlation with agitation, we would expect). We thus designed our experiment to also take form-emotion correspondences into account. The corresponding null hypothesis would be that the associations between specific colors/forms and specific emotions would be statistically unreliable or random.
Experimental Design We constructed an online survey with passwordprotected login and automatically tabulated results. The survey contained four sections. Section one consisted of a participation-consent form in keeping with IRB (Institutional Review Board) guidelines for research involving human subjects. Section two invited participants to provide simple demographic information, age, gender, and state or country of origin.
Section three presented 21 abstract figures, each approximately 2 x 3 inches in display size, each presenting a distinct form/color combination. Figures sampled the whole range of form possibilities (varietycontrast-pattern) and color temperatures (red-orangeyellow-green-blue-purple). Figures consisted of highresolution .jpg files, checked in several web browsers and PC platforms for color fidelity. Participants were asked to make three choices from a menu of twelve possible emotional responses, in alphabetical order: A. agitated B. amused C. attracted D. calmed E. challenged F. concerned G. diverted H. focused I. organized J. resolved K. rested L. stimulated The alphabetical order of the menu options served to scramble our postulated emotion spectrum and reduce the likelihood of order-of-mention bias. Including twelve options and three choices among them allows for the possibility of nuanced gradations between the six major divisions predicted by the Peircean model, also allowing us to refine our emotion terminology, to discover whether some emotion terms were apt than others based on the experimental results. Section four repeated 10 of the figures from section three, specifically the non-colored forms. Participants were offered a menu of twelve HTML-defined colors (this time in spectral order, red-orange-yellow, etc.) and asked to pick three colors with which the abstract black and white forms seemed to best correspond. Technical communication students enrolled at a regional university were invited to participate in this survey. They received no instruction or information about form/color/emotion correspondences prior to the study. Participation was strictly voluntary. Of 44 invited participants, 33 successfully completed the survey.
Results: form/color and emotion spectra From the compiled responses we can generate a histogram for each survey figure. Each histogram can be thought of as the spectral emotion analysis of its figure, directly analogous to the spectrum of color components that can be extracted from any given light source, or the wave components extracted from any given sound source. In most cases, the spectral response to each figure is consistent with our hypotheses, correlations of form/color and emotion are statistically reliable, at least in this population of survey participants. The null hypothesis is decisively ruled out by the data, that only essentially random and idiosyncratic associations would be found between form/color and emotion. Figures 1, 2, 3, and 4 illustrate general patterns found in the data. Broadly speaking, form has a more powerful effect on perceived emotion, but different color hues also make distinct, non-negligible contributions to overall emotional response.
Figure 2. diverted-rested-calm range responses from yellow-green-blue range color and/or variablenoncontrastive form. Figure 1. Agitated-diverted range responses from redyellow range color and/or contrastive- variable form. Figure 1 shows the range of form and color that evoke a predominantly challenged-agitated-diverted response and predominantly suppress the range of rested-calmedorganized responses. Note that the intensity of this effect is strongest in the black-and-white form (peak response = 21), and weakest for the solid color field (peak response = 14). Even the solid color response is statistically significant however (standard deviation = 5.8, mean response = 8). Figure 2 shows the range of form and color that evoke a predominantly diverted-rested-calm response and suppress challenged-agitated-diverted responses. Again the effect is most strongly tied to form, specifically the absence of strong contrast. As a result, the solid (i.e. utterly non-contrastive) color field boosts the calming effect (peak response = 27) but also the solid color also suppresses the diverted-amused effect, presumably due to a lack of variable form. Figure 3 shows the range of form and color that evoke a predominantly calm-organized-focused response and suppress agitated-diverted-rested responses. Here the repetition of form elements (i.e. pattern) accounts for the largest effect (peak = 22), but the effect remains significant even in the minimally patterned colored form.
Figure 3. Pattern and blue-range color effects.
Figure 5. Form-color associations mediation of emotional response.
Figure 4. The "calm ceiling" for unpatterned blue vs. "calm cancellation" for high-contrast green. Figure 4 illustrates further form-over-color effects. In particular, blue-range color does not readily evoke a feeling of organization or focus without at least minimal formal pattern support (compare Figure 3 bottom). Thus, unpatterned blue color apparently hits the "calm" ceiling, indicating a range of emotion to which color may contribute, but only with the "boost" of formal patterning. Conversely, the addition of heavy formal contrast and variety to an otherwise "restful" color like green effectively cancels any restful/calming effect. Figure 5 samples results from section four of the survey. Here we investigate whether formal patterns can evoke a consistent sense of color directly, or whether form-color associations are essentially mediated by emotional responses. The results illustrated suggest that there is a form/color association roughly parallel to form/emotion and color/emotion associations, comparing the raw color responses in Figure 5 with the emotion spectra for the black-and-white stimuli in Figures 1, 2, and 3. However, the entire body of data suggests only a strong correlation (0.71) for Red-range color associations with contrastive form. Yellow-range associations are not found in this data. Green range associations with formal variety are present but weak (0.44) and Blue-range associations with pattern are weaker still (0.18). Overall results are shown in Table 2.
Table 2. Combined data for form-color association without emotion-response mediation.
Discussion: theory and results Visual design elements have elsewhere been classified within a Peircean framework as decorative if their primary purpose is to evoke feeling or in other words emotion [10, p.2]. A decorative by definition in this framework is perceived as a unified gestalt, is interpreted by means of association to prior perceptual experience, and is interpreted as feeling rather than as direct reference to physical things or to information.[10, p.5].
 Oyama, T., Affective and Symbolic Meanings of Color and Form: Experimental Psychological Approaches, Empirical Studies of the Arts, 21: 137-142, 2003.  Osgood, C.E., Saci, G.J., & Tannenbaum, P.H., The Measurement of Meaning, University of Illinois Press, Urbana IL, 1957.  Puhalla, D.M., Perceiving Hierarchy Through Intrinsic Color Structure, Visual Communication, 7:199-228, 2008.
Conclusion: results and usability
 Peirce, C.S. Collected papers, vol. 1-6, ed. Hartshorne, C. & Weiss, P., vol.7-8, ed. Burks, A., Harvard Univ. Press, Cambridge, 1935, 1958.
All in all this turned out really well. One can envision emotion-spectrum analysis being part of the usability test of any decoration-heavy visual design.
 Amare, N. and Manning, A., A Language for Visuals: Design, Purpose, Usability, IPCC Conference Proceedings, IEEE, Piscataway, NJ, 2008.
References  Kress, G. and Van Leeuwen, T., Colour as a semiotic mode: notes for a grammar of colour, Visual Communication, 1: 343 - 368, 2002.  Collier, G.L., Affective Synesthesia: Extracting Emotion Space from Simple Perceptual Stimuli, Motivation and Emotion, 20: 1-32, 1996.  Zentner, M.R., Preferences for colours and colour-emotion combinations in early childhood, Developmental Science 4: 389 - 398, 2001.  Ou, L., Luo, M.R., Woodcock, A., Wright, A., A Study of Colour Emotion and Colour Preference. Part I: Colour Emotions for Single Colours, COLOR Research and Application, 29: 232-240, 2004.  Xin, J.H., Cheng, K.M., Taylor, G., Sato, T., Hansuebsai, A., Cross-Regional Comparison of Colour Emotions Part II: Qualitative Analysis, COLOR Research and Application, 29: 458-466, 2004.
About the Authors Alan Manning is a professor of linguistics and English language at Brigham Young University. He teaches graduate courses in writing and research design and undergraduate courses in linguistics and editing. He is a coauthor of Revising Professional Writing Science and Technology, Business, and the Social Sciences (with Riley, Campbell, and Parker, Parlay Press). Nicole Amare is an assistant professor of technical writing at the University of South Alabama. She teaches technical writing, editing, stylistics, and grammar. She has written Real Life University, a college success guide, and has edited Global Student Entrepreneurs and Beyond the Lemonade Stand. With Barry Nowlin, she is a coauthor of Technical Editing: Products and Processes (forthcoming from Pearson Education).