Haptic and Tactile Adjectives Are Consistently Mapped onto Color Space

Multisensory Research (2015) DOI:10.1163/22134808-00002512 brill.com/msr

Haptic and Tactile Adjectives Are Consistently Mappedonto Color Space

Yasmina Jraissati 1,∗, Nadiya Slobodenyuk 2, Ali Kanso 3,

Lama Ghanem 2 and Imad Elhajj 3

1 Department of Philosophy, American University of Beirut, Lebanon2 Department of Psychology, American University of Beirut, Lebanon

3 Department of Electrical and Computer Engineering,American University of Beirut, Lebanon

Received 22 February 2015; accepted 17 July 2015

AbstractCross-modal associations refer to non-arbitrary associations of features across sensory modalities.Such associations have been observed between many different sensory features. One association thathas rarely been studied so far is between touch and color. In this study, participants were asked tomatch tactile and haptic adjectives to color samples shown individually on a screen. They couldselect one to 11 tactile and haptic terms, presented in 11 pairs of opposed adjectives. The resultsshowed a regular pattern in the way tactile and haptic terms were matched to color. Our resultsfurther revealed that the colors to which tactile and haptic terms were matched did not fall within theboundaries of color lexical categories, suggesting that the associations were not based on lexicon —despite the frequent occurrence of linguistic expressions such as ‘soft pink’, not all colors called‘pink’ were matched to ‘soft’. In contrast with one recent study, the distribution of tactile and hapticterms across the Munsell array suggests that along with brightness and chroma, hue was also relevantto participants’ responses. Specifically in the case of hue, several opposed adjectives were relativelywell matched to opposed colors, along the orthogonal Yellow/Blue and Red/Green axes, which aresuggested to structure the space of hue experience. Possible accounts of these results are considered.

KeywordsCross-modal associations, color, color lexicon, touch, haptic lexicon, tactile lexicon

* To whom correspondence should be addressed. E-mail: [email protected]

© Koninklijke Brill NV, Leiden, 2015 DOI:10.1163/22134808-00002512

2 Y. Jraissati et al. / Multisensory Research (2015)

1. Introduction

In addition to the arguably most well known and robustly established ‘Kiki–Bouba effect’ that refers to the associations between the sound ‘Bouba’ anda rounded shape and the sound ‘Kiki’ and a jaded, angular shape (Brem-ner et al., 2013; Köhler, 1929; Ramachandran and Hubbard, 2001), manyother associations across different sensory modalities have been reported aswell. Systematic matches have been observed between brightness and loud-ness (Marks, 1974); flavor and sound (Crisinel and Spence, 2010), musicaltimber and smell (Crisinel et al., 2012; Deroy et al., 2013), and brightnessand pitch (Marks, 1974; Walker et al., 2010). Color was found to be system-atically associated with music (Palmer et al., 2013), taste (Tomasik-Krotkiand Strojny, 2008), and smell (Deroy and Spence, in press, for a review;Kim, 2013; Schifferstein and Tanudjaja, 2004). When it comes to tactile ex-perience, associations between color and temperature (Ho et al., 2014), colorand weight (Monroe, 1925), and lightness and vibro-tactile frequency (Mar-tino and Marks, 1999) have been reported. Overall, however, the associationsbetween color and touch have been largely under-researched until recently(Ludwig and Simner, 2013). The current study intends to fill this gap in theliterature by identifying the role of hue in cross-modal associations involvingcolor and providing a novel, easily replicable paradigm.

In a recent study of the touch/color interface, Ludwig and Simner (2013)presented participants with objects varying in degrees of smoothness-to-roughness, softness-to-hardness, and pointiness-to-hardness, and asked themto match these sensations to a color by selecting it from a color wheeldisplayed on the screen. Ludwig and Simner found systematic associationsto brightness and chroma: softer and smoother stimuli were systematicallymatched to higher values of brightness and chroma, while harder and rougherstimuli were systematically matched to lower values of brightness and chroma.Rounder stimuli were matched to higher values of brightness than pointierstimuli, but no effect was observed in chroma. No systematic match to huewas found.

An important difference between brightness, chroma (or saturation), andhue should be noted. Brightness and chroma vary along a single linear di-mension: a given color is more or less bright; it is more or less saturated.Specifically, the linearity of brightness and chroma imply that they start at apoint of zero brightness (black), and zero chroma (white, black, or grey), grad-ually and continuously increasing from there. On the other hand, hue does notvary in this way. Although hue is also represented as continuously varyingalong one dimension, this dimension is not linear, but circular. There is nozero starting point in hue, like there is in brightness and chroma. If a hue is‘less red’ than another, it could either be yellowish or bluish. There is no such

Multisensory Research (2015) DOI:10.1163/22134808-00002512 3

ambiguity in brightness or chroma, when a color is said to be less bright orless saturated than another.

For this reason, Ludwig and Simner (2013) could not examine the role ofhue in the association to tactile sensations in the way they did with bright-ness and chroma. In order to assess whether hue, along with brightness andchroma, was matched to tactile sensations, Ludwig and Simner categorizedparticipants’ color responses into 11 so-called basic color terms (Berlin andKay, 1969). In other words, they resorted to the way these hues can be named.They found that tactile sensations did not match color categories. For example,they report that only bright, not dark, shades of what we call ‘pink’ in Englishare associated with soft sensations, while only dark (and not bright) shades ofwhat we call ‘brown’ are associated with rough sensations. They concludedfrom such observations that tactile sensations are not systematically matchedto hue. However, due to the reduction of hue responses to lexical categories,this conclusion should be reframed as an absence of systematic associationsbetween tactile sensations and lexical color categories.

Although the use of such reduction of responses to lexical color categoriesmight have made answering the question of the role of hue in Ludwig andSimner’s experiment more difficult, the study of the role of color lexical cat-egories in cross-modal associations involving color is legitimate. When usinglanguage, we seem more inclined to associate some tactile adjectives to somecolor terms rather than others. Take the English color terms ‘pink’ and ‘brown’for example. In English, there are more occurrences of ‘soft pink’ (8.9E-05%of all words in all English books available online in the year 2000) than ‘roughpink’ (1.8E-07%), and there are more occurrences of ‘rough brown’ (1.4E-06%) than ‘rough pink’ (‘Google books Ngram Viewer’, n.d.). Why is that thecase? Amongst the mechanisms that are thought to mediate cross-modal asso-ciations, language was suggested as a possibility. For example, ‘high’ refers toboth the elevation in space and the elevation of pitch. The fact that both audi-tory and spatial attributes are labeled ‘high’ is considered a possible cause ofsuch associations (Gallace and Spence, 2006; Karwoski et al., 1942; Melaraand Marks, 1990; Spence, 2011; Walker, 2012). If the association of ‘soft’ to‘pink’ is exclusively lexical, we would expect all color shades labeled ‘pink’to be called ‘soft’ to the same degree. However, this does not seem to be thecase.

With regard to the role of sensory lexicon, Ludwig and Simner’s (2013)study provides insight into one particular issue — the match between tactilesensations and lexical color categories. The fact that no such systematic matchwas found means that even if ‘soft’ often occurs in association to ‘pink’ inlanguage, we should not expect all shades included in the extension of theterm ‘pink’ to be associated to ‘soft’ to the same degree; or, more generally,we should not expect ‘soft’ to be matched to shades of color on the basis of

4 Y. Jraissati et al. / Multisensory Research (2015)

their label. If the term ‘soft’ does not map onto the entire extension of thelexical color category ‘pink’, what characterizes the color shades to which‘soft’ matches? How is the term ‘soft’ distributed in color space? Based onLudwig and Simner’s findings we would expect ‘soft’ to be matched to brightcolors. However, whether ‘soft’ and other tactile terms would be matched tohue as well remains an open question.

The analysis of Ludwig and Simner’s (2013) paradigm and findings alsoraises the question whether such associations could be found if the paradigmwere modified to prompt the match of haptic and tactile terms to color, ratherthan haptic and tactile sensations. Such a paradigm change would also ensurethat the associations between touch and color are not lexical. If haptic andtactile terms do not distribute onto the given color array (Munsell, 1941), ac-cording to the color lexicon boundaries, then the possibility of lexical matchwould be overruled.

Another consideration for the development of an optimal paradigm to studythe role of hue pertains to the direction of the matching task. Asking partici-pants to match haptic terms to color sensations in the Munsell two-dimensionalarray is likely to bias the way in which related or opposed terms become dis-tributed on the array due to memory bias. Instead, if participants are asked tomatch individual color samples presented at random with haptic and tactileterms, the chance that they would remember what color they had matched towhich terms is low.

Taking into account the above considerations we asked participants tomatch color samples presented individually on a screen to haptic and tactileterms. The terms they were presented with referred both to tactile propertiesof surfaces (for example smooth/rough, hard/soft, warm/cold) and to hap-tic properties resulting from active interaction with the object (for examplelight/heavy, elastic/inelastic). We expected to find a systematic distribution ofhaptic terms across color space. For example, we expected the highest consen-sus for the ‘rough’ and ‘hard’ matches to be concentrated in the darker, thuslower, parts of the Munsell array used in this experiment, knowing that in theMunsell array, brightness levels vary along the y-axis, with black at the bottomand white at the top (more on the Munsell system below). We also expectedthe respectively opposed ‘smooth’ and ‘soft’ terms to be most consensuallymatched to the brighter, thus higher, parts of the Munsell array. In the caseof chroma, our color stimuli and experimental setup being different from thatof Ludwig and Simner’s (2013), it was not clear whether we would observethe same pattern of responses. Finally, when it comes to the way haptic termswould be matched to hue, previous results by Ludwig and Simner did not al-low us to make specific predictions. Thus, either tactile and haptic terms arenot matched to hue, and therefore haptic terms would be distributed across theMunsell array exclusively along the single dimension of brightness, or these

Multisensory Research (2015) DOI:10.1163/22134808-00002512 5

terms are also matched to hue, and terms like ‘rough’ and ‘hard’ would notonly be most consensually matched to the darker, lower, parts of the Munsellarray, but also, within this dark part of the Munsell array, the highest con-sensus for ‘rough’ and ‘hard’ would be concentrated in one area, towards theleft, center, or right, of the array, knowing that hue varies along the x-axis inthe Munsell array, ranging from red to the left, passing by yellow, green, bluetowards the center, then purple and red again towards the right of the array.Therefore, if tactile and haptic terms are matched to both brightness and hue,the distribution of the terms in color space should depend on both dimensionsof brightness and hue in the two-dimensional Munsell array.

In short, this study was designed to answer three questions: (1) Are tactileand haptic terms matched to color terms in color space? (2) If not, as we expectbased on Ludwig and Simner (2013): Are haptic terms matched to color sen-sation systematically? More specifically, if they are systematically distributed,are they matched to hue, as well as to chroma and brightness? Finally (3) is theproposed experimental paradigm successful in examining the role of hue? Ifyes, this study would also provide a replicable empirical paradigm that wouldallow assessing the role of hue in studies of cross-modal associations involvingsensory vocabulary.

As expected, we found that tactile and haptic terms are not matched tolexical color categories. However, haptic terms are matched to color in a sys-tematic way, in chroma, brightness and hue dimensions. Furthermore, twofindings exceeded our expectations. First, we found that in the case of someterms, and in line with prototype theory (Rosch, 1999), there is a central colorthat is most consensually matched to ‘rough’, ‘hard’, etc., and consensus de-creased gradually away from this central sample in a radial manner. Second,opposed haptic adjectives seemed to oppose in color space. In what follows,we present the study and discuss these findings.

2. Methods

2.1. Participants

Fifty-nine students completed the experiment for a research participationcredit in the Introduction to Psychology course at the American Universityof Beirut. All participants had normal or corrected-to-normal vision as wellas normal color vision. Ishihara plates were used to test for normal colorvision prior to the experiment. The study was conducted in Arabic and par-ticipants’ consideration of Arabic as their native language and the expectedfluency served as inclusion criteria. The study was approved by the Institu-tional Review Board (IRB) of the American University of Beirut.

6 Y. Jraissati et al. / Multisensory Research (2015)

2.2. Stimuli and Design

2.2.1. Color StimuliThe present research used the Munsell color system. The use of Munsell wasmotivated by our intention to compare distribution of haptic terms in colorspace to that of color terms. A Lebanese Arabic lexicon using the WorldColor Survey protocol (WCS, http://www1.icsi.berkeley.edu/wcs/) was pre-viously obtained using the Munsell array (Jraissati, 2010), and a lot of dataon color naming using Munsell is also available for possible future com-parisons (http://www1.icsi.berkeley.edu/wcs/data.html). The Munsell systemrepresents a three-dimensional color space with the axes being hue, chromaand value. The Munsell color system was developed for surface colors, in pa-per. It is characterized by perceptual uniformity in that the perceptual distancebetween one pair of colors and the adjacent pair of colors in each of the threedimensions is taken to be the same (Newhall et al., 1943). The space is notperfectly spherical as the levels of saturation reached by each hue at differ-ent levels of lightness vary and depend upon available technology. Sixty-foursamples from the surface of the Munsell solid were chosen for this study.A Mercator projection of the Munsell solid’s surface consists of an array ofthe 320 most saturated color shades of the model. Therefore, the color sam-ples chosen for the present study were the most saturated, with chroma varyingacross samples. Given the fact that we wanted to compare distribution of hap-tic terms with that of color terms, only the surface of the Munsell solid wasused, as in WCS color naming studies. In the array, the 40 available hues arerepresented horizontally on the x-axis, each column corresponding to a huebeing attributed a number (from 1 to 40). The eight available levels of bright-ness, or values in Munsell notation, are represented vertically on the y-axis,each row corresponding to a value being attributed a letter (from B at the topto I at the bottom, with A for white, and J for black). The 64 samples usedin this experiment (60 chromatic shades, and four achromatic shades) wereregularly spaced on the array. They were picked so as to represent the wholearray, and not favor one area of the space over another. Twenty hues out of the40 were selected, ranging in a regular pattern across the array. If one hue wasincluded, the next was excluded, thus yielding the following hue selection:2.5R, 7.5R, 2.5YR, 7.5YR, 2.5Y, 7.5Y, 2.5GY, 7.5GY, 2.5G, 7.5G, 2.5BG,7.5BG, 2.5B, 7.5B, 2.5BP, 7.5BP, 2.5P, 7.5P, 2.5RP, 7.5RP. Three values werechosen for each hue, across the 8 available values (2 to 9), excluding levels0-black, 10-white, and 1. The exclusion of level 1 is consistent with the WorldColor Survey color lexicon studies (Cook et al., 2005). In order to ensurethat both the minimum value (2) and the maximum (9) were equally repre-sented in our color set, the values alternated across the selected hues. Thus, at2.5R, the three values were 3, 6, and 9, while at the next selected hue 7.5R,

Multisensory Research (2015) DOI:10.1163/22134808-00002512 7

the three values were 2, 5, and 8. This alternate pattern was repeated overthe 20 hues. Finally, four achromatic shades were added to the 60 chromaticshades: black (brightness level 0), grey (4), grey (7), and white (10), yieldingfour equally spaced achromatic shades. In order to obtain sRGB values for theselected Munsell colors that were presented on screen, we referred to a listprovided by color scientist Harold Van Aken, and published on his website(http://www.wallkillcolor.com/muntable.csv). The conversion assumed an Il-luminant C, and two-degree observer. In the absence of a spectrophotometer,we could not measure the colors rendered on the screen. We relied on the factthat the selected color samples were quite spaced (five Munsell hue steps andthree Munsell value steps) across the array to guarantee that they would beexperienced as different. This was enough for the purpose of this study. Noneof the Munsell colors converted into sRGB in this study were out of gamut.

2.2.2. Tactile and Haptic LexiconTen pairs of adjectives that pertain to substance properties and one that pertainsto shape were selected. Shape adjectives were added on the basis of Ludwigand Simner’s (2013) study that included tactile stimuli pertaining to mate-rial properties (rough/smooth and hard/soft) as well as shape (pointed/round).In the literature, rough/smooth, hard/soft, dry/wet, thin/thick were identifiedas possible dimensions of tactile space, along with a possible dimension re-ferring to slipperiness (see Guest et al., 2011, for a review). In this study,apart from including terms that referred to the sensations used in Ludwig andSimner, haptic and tactile adjectives were selected so as to cover the widestpossible range of haptic and tactile sensations. Thus, to the five dimensionsdiscussed in Guest et al. (2011), and to pointy/round that was included inLudwig and Simner (2013), we added terms referring to the haptic sensationof weight (light/heavy) and the tactile sensation of temperature (warm/cold).Both weight (Monroe, 1925) and temperature (Ho et al., 2014) have beenfound to be associated to color in the past.

Haptic adjectives were in opposed pairs: (1) smooth/rough; (2) soft/hard;(3) sticky/non-sticky; (4) supple/rigid; (5) elastic/stiff; (6) viscous/fluid;(7) light/heavy; (8) warm/cold; (9) thin/thick; (10) dry/humid; and (11) pointy/round. In Arabic: (1) meles/khechen; (2) na’im/qasi; (3) laseq/ghayr laseq;(4) layyen/jamed; (5) mutamaghet/yabes; (6) lizij/sayel; (7) khafif/thaqil;(8) dafi/bared; (9) rafi’/samik; (10) nashef/roteb; (11) murawwas/mudawwar.In what follows, Arabic haptic terms are referred to in English.

2.2.3. Linguistic ProficiencyLebanon is a multiethnic Arab country with a large percentage of multilingualpopulation, and many considering their first language to be Arabic, French,English, or Armenian. To ensure that Arabic was the first language of theselected group of participants and that they share the same level of Arabic

8 Y. Jraissati et al. / Multisensory Research (2015)

language proficiency, the L2 Language History Questionnaire developed byBrain, Language, and Computation Lab at Penn State University was used (Liet al., 2014).

2.2.4. ProcedureParticipants were seated at a distance of 50 cm from a 17′′ computerscreen (Fujitsu Siemens Computers GmbH, LCD COLOR MONITOR,SCENICVIEW B17-5) with a resolution of 1280 by 1024 pixels. The screendepicted one color sample on a grey background and 11 pairs of adjectives.

Participants were presented with one color sample at a time and in a randomorder. Below each color sample, the 11 pairs of opposed adjectives were listed.The opposed haptic terms were presented each on one end of a continuumthat depicted a slider in the middle. They were instructed to match the hapticadjectives to the presented color by moving the slider from the central point toone of the two sides, thereby choosing an intensity of the haptic adjective. Thecloser to the adjective the slider was moved, the more of that characteristic theparticipant was choosing. Discrete divisions of the slider were used to recordthe response even though the slider looked continuous to the participants. Eachhaptic adjective could receive a value of 1 to 5, with value 5 representing themaximum of the characteristic chosen by the participants. Once the participantwas seated in front of the screen, the experimenter orally instructed him asfollows: “You will be presented with 64 different color patches. Please choosefrom the terms below the ones that you feel match these colors. You have tochoose at least one adjective, and up to 11”. Participants were instructed notto overthink their responses. The experimenter then read aloud the list of alltactile and haptic terms, starting from right to left. Each color sample waspresented one single time.

3. Results

3.1. The Use of Tactile and Haptic Adjectives

3.1.1. Overall Frequency of Tactile and Haptic AdjectivesParticipants were allowed to choose as many adjectives as they wished for eachcolor sample, and required to choose at least one. Considering that there were11 pairs of opposing adjectives, the largest number of adjectives that couldhave been chosen for each color sample was 11. On average, participants chose5.36 (SD = 3.29) haptic adjectives per color sample. The analysis of thesechoices revealed that some haptic adjectives were used to describe many colorsamples, while others were used infrequently. The most frequently used adjec-tives across the color samples were ‘soft’ and ‘cold’; both with the highest rateof achieved participant consensus of 78%. The participant consensus refers tothe percentage of participants who have chosen a specific haptic adjective for

Multisensory Research (2015) DOI:10.1163/22134808-00002512 9

a particular color sample. Thereafter we use the term ‘frequency’ of haptic ad-jectives and participant ‘consensus rate’ interchangeably. The least frequentlyused adjectives across color samples were ‘elastic’ and ‘pointy’ with highestrates of achieved consensus of 28.8% and 29.3% respectively. Each term, ex-cept for ‘rough’, was used at least by one participant for all color samples (fulldata are available online in Supplement 1).

3.1.2. Tactile and Haptic Adjectives with the Highest Participant ConsensusThere was also an overall high participant agreement on the ‘best fitting’ hapticadjectives for specific color samples. Best fitting adjectives were consideredthe ones that received the highest participant consensus for a specific colorsample. Ten color samples had at least one haptic term with a consensus rateover 70%; 15 color samples had at least one haptic term with a consensus over60%. An overall of 47 color samples, out of 64 used in the study, had at leastone haptic term with a consensus rate over 50%, and only five color samplesout of 64 had no haptic terms with consensus rates over 39%; these were: D0(an achromatic grey, of brightness value 7) E17 (hue 2.5G, value 6, in Munsellnotation), F15 (hue 7.5GY, value 5), H21 (hue 2.5BG, value 3) and C15 (hue7.5GY, value 8).

There was a difference in the number of color samples a specific adjec-tive was considered best fitting for. The adjective ‘heavy’ received the highestparticipant consensus for the largest number of color samples. ‘Heavy’ wasconsidered best fitting for 21 color samples, followed by ‘cold’ (best fittingfor 13 color samples), ‘soft’ (10 color samples), ‘warm’ (seven color sam-ples), ‘light’ (five color samples), and ‘viscous’ (four color samples). ‘Dry’,‘smooth’, ‘hard’, and ‘fluid’ were each considered best fitting for only onecolor sample, and the rest of the haptic adjectives did not have the highestparticipant consensus for any of the color samples.

3.1.3. Tactile and Haptic Adjectives, and Arabic Color LexiconIn order to identify whether tactile and haptic terms are matched to color termsin color space, we needed to compare the distribution of haptic terms in colorspace to that of color terms. A Lebanese Arabic color lexicon was obtained byJraissati (2010), using the World Color Survey (WCS) protocol (Cook et al.,2005). The Lebanese partitioning of color space obtained by Jraissati (2010)was similar to the one previously obtained for Lebanese by Berlin and Kay(Berlin and Kay, 1969), except that in Jraissati (2010) the terms most fre-quently used for ‘orange’ and ‘purple’ were borrowed from French (‘orange’and ‘mauve’ respectively) instead of the Arabic ‘burtuqali’ and ‘lailaki’ re-ported by Berlin and Kay (1969). According to Jraissati (2010), the basic colorterms for the Lebanese color lexicon are: ‘abyad’ (‘white’), ‘aswad’ (‘black’),‘ahmar’ (‘red’), ‘asfar’ (‘yellow’), ‘azraq’ (‘blue’), ‘akhdar’ (‘green’), ‘benni’(‘brown’), ‘rmedi’ (‘grey’), ‘zahri’ (‘pink’), ‘orange’ (‘orange’), and ‘mauve’

10 Y. Jraissati et al. / Multisensory Research (2015)

(‘purple’). In what follows, we refer to the Arabic color terms using their En-glish equivalents.

The comparison of the distribution of haptic terms in color space to that ofcolor terms using the Lebanese Arabic color lexicon is given in Fig. 1A andB. Taking the warm/cold and soft/hard pairs as an example, Fig. 1A showsthat only 10% of colors matched to ‘warm’ are in the extension of ‘red’; therest is distributed across the extensions of ‘white’, ‘pink’, ‘purple’, ‘orange’,‘yellow’, and ‘brown’. Of colors matched to ‘cold’, 66.6% are in the extensionof ‘blue’; the rest is distributed across the extensions of ‘white’, ‘purple’, and‘green’. Similarly, 41.2% of colors matched to ‘soft’ are also named ‘pink’,with the rest of ‘soft’ being distributed across ‘white’, ‘purple’, ‘green’ and‘blue’. Hundred percent of colors named ‘hard’ are also named ‘black’. How-ever, and conversely, only 40% of the colors named ‘black’ are matched to‘hard’ (Fig. 1B). In other words, the remaining 60% of colors named ‘black’are not matched to ‘hard’. Similarly, only 35% of colors named ‘pink’ arematched to ‘soft’. Also, 16.6% of colors named ‘red’ are matched to ‘warm’,and only 26.6% of colors named ‘blue’ are matched to ‘cold’. This shows thattactile and haptic terms are distributed over the extensions of several colorterms, and conversely, not all the colors included in the extension of a lexicalcolor category are matched to a given tactile or haptic term.

In order to examine the relationship between haptic terms, color terms, andthe color samples of the Munsell array, as well as offer a visual representationof this relationship, we resorted to the methodology mostly used in the colorcategorization literature, namely a ‘mode map’ of the Arabic color lexicon,and ‘term maps’ of haptic and tactile terms. In color categorization, a ‘modemap’, is a visual representation of the distribution of color terms in the Munsellarray, where the term used most frequently in reference to a given color isassociated to that color in the map. As a result, the average partitioning of thecolor space by a given color lexicon is visualized. ‘Term maps’ by contrast,consist of a visual representation of the distribution of one specific term acrossthe Munsell array, including the color samples to which the term is not mostfrequently associated. Term maps provide an insight into the internal structureof the term’s extension beyond the lexical boundaries of the mode map. In thisstudy, in the term maps of warm/cold and soft/hard that are used as example,we opted for a cut-off point of 50%.

For illustration, here we present in Fig. 2A and B the term maps ofwarm/cold and soft/hard, respectively, over the average partitioning by theArabic lexicon of the Munsell array (mode map). The term maps confirm thatthe distribution of color terms and tactile and haptic terms is different: the col-ors matched to ‘cold’, ‘warm’, ‘soft’ and ‘hard’, range over color space acrossthe extensions of several lexical color categories. Also, not all colors carry-ing a given color label are matched to a tactile or haptic adjective to the same

Multisensory Research (2015) DOI:10.1163/22134808-00002512 11

Figure 1. In (A) the bars indicate, among the color samples matched to a tactile/haptic term,the percentage of those which carried the label of a given color category. For example: 41.2%of colors matched to ‘soft’ are labeled ‘pink’. Conversely, in (B) the bars indicate, among thecolor samples carrying a certain color label, the percentage of those which were matched to agiven tactile/haptic term. For example 35% of colors labeled ‘pink’ are matched to ‘soft’. Thecut-off point for both tactile/haptic and color lexicons is 50%.

12 Y. Jraissati et al. / Multisensory Research (2015)

Figure 2. Distribution of the adjectives cold/warm (A) and soft/hard (B) onto the Munsell colorarray as partitioned by the Lebanese Arabic color lexicon (mode map). The letters C and Windicate the location of the adjectives ‘cold’ and ‘warm’, respectively, on the Munsell array,while S and H stand for ‘soft’ and ‘hard’. The size of the letters indicates the strength of theassociation (i.e., participant consensus as measured by the frequency of choice in a group ofparticipants) between a target color sample and an adjective ‘cold’, ‘warm’, ‘soft’ and ‘hard’.The largest letters indicates the frequency 65 � x; medium sized letters indicate the frequency50 � x < 65%; smallest sized letters, in small caps, indicate the frequency 25 � x < 50% andwere included here to give a sense of the wider extension of the haptic term in color space.The blocks stand for different Arabic color terms. From left to right, bottom to top: ‘white’,‘grey’, ‘black’, ‘pink’, ‘red’, ‘orange’, ‘brown’, ‘yellow’, ‘green’, ‘blue’, ‘purple’. This figureis published in color in the online version.

degree (all term maps of tactile and haptic adjectives are available online inSupplement 2).

3.1.4. Radial Structure of Tactile and Haptic Adjectives in Color SpaceTerm maps (as in Fig. 2A and B) suggest that some haptic and tactile terms fea-ture a radial structure in the way they cluster in color space. Radial structure ofthe tactile and haptic terms in the present context refers to the extension of theterms in color space, such that the category includes a highly consensual cen-tral part and a peripheral part with a lower participant consensus, and the levelof consensus decreases from the center towards the boundaries of the hapticcategory in a gradual manner. In the given example (Fig. 3) of smooth/rough,the reached level of consensus is maximal in value 9 for ‘smooth’, while it ismaximal in value 2 for ‘rough’. Within each value, some neighboring hues aremost consensually matched to ‘smooth’ or ‘rough’, and as we move away from

Multisensory Research (2015) DOI:10.1163/22134808-00002512 13

Figure 3. Histograms of the frequency of match of smooth/rough to hue, presented along theMunsell hue circle for values 9 (top) and 2 (bottom). On the circle, the hues vary clockwise fromR (at the top of the circle) to Y, YR, Y, GY, G, BG, B, PB, P, RP, and include the four availablesteps for each hue (2.5, 5, 7.5 and 10) knowing that only two steps were used in this study(2.5 and 7.5). The concentric circles represent the reached consensus level, with increment of5%. Thus for example, at brightness level 9, smooth is matched to the hue 2.5R by x% of theparticipants, where x is 60 < x � 65%.

these most consensually matched hues, the level of consensus decreases. Thisbehavior is characteristic to radial structure. In Fig. 3, histograms of reachedlevels of consensus for a given color (at a given hue and value) are presentedon the Munsell hue circle and are therefore circular.

Maps of the consensus rates of the adjectives in color space suggest thatseveral haptic and tactile terms feature a radial structure: consensus decreasesclearly in a gradual manner from the center of the category to its periphery inboth the hue and value dimensions of the Munsell array. The haptic terms withradial structure are: ‘smooth’ (reached level of ‘consensus’ 62.7% at value 9),‘rough’ (55.9% at value 2), ‘soft’ (78% at value 8), ‘hard’ (69.5% at value 2),

14 Y. Jraissati et al. / Multisensory Research (2015)

‘light’ (71.2% at value 9), ‘heavy’ (76.3% at value 2), ‘warm’ (70.7% at value2), ‘cold’ (78% at value 6), ‘humid’ (71.2% at value 6). On the other hand,the remaining adjectives do not feature such radial structure clearly; they ei-ther do not have a consensual center, or the change with regard to the mostconsensual sample is either not clearly gradual or non-gradual. Several hapticterms reach high consensus on a few color samples but do not feature a cleargradual structure. ‘Thin’, for example, has two most consensual color samplesreaching 49.2% (2.5BG and 2.5RP), both at value 9, but consensus does notdecrease in a gradual manner in any dimension. ‘Rigid’ has a highly consen-sual center (55.9% at value 2), but the extension of this haptic category is notclearly gradual, with consensus decreasing and increasing again in a seem-ingly arbitrary way. The same is true for ‘stiff’ (62.7% at value 2), ‘viscous’(52.5% at value 3), ‘thick’ (54.2% at value 0), ‘supple’ (49.2%, at value 8).The radial structure of some categories rather than others is just another in-dicator of the fact that matches of color to these haptic/tactile terms was notarbitrary, but guided by similarity relations between colors (all term maps areavailable at: http://jraissati.com/tactile-adjectives-data/).

3.2. Brightness, Chroma and Tactile and Haptic Adjectives

Brightness and chroma of the color samples that received the ratings over40% of the opposed haptic terms were compared using paired-samples t-testsfor the following pairs: smooth/rough, soft/hard, supple/rigid, light/heavy,warm/cold, thin/thick, and dry/humid. A lower consensus threshold (40%)than in other analyses was chosen to ensure sufficient number of color samplesper pair of terms.

Smooth colors (M = 6.57, SD = 1.72) were brighter than rough colors(M = 2.57, SD = 0.98), t (6) = 4.58, p < 0.01, and had higher chroma value(M = 7.71, SD = 4.07; M = 3.14, SD = 2.79, respectively), t (6) = 2.64,p < 0.05.

Soft colors (M = 6.56, SD = 2.07) were brighter than hard colors (M =2.56, SD = 1.13), t (8) = 4.71, p < 0.01, but did not differ on chroma, t (8) =1.57, p > 0.05.

Supple colors (M = 7.33, SD = 1.51) were brighter than rigid colors (M =2.50, SD = 1.38), t (5) = 6.87, p < 0.01, but did not differ on chroma, t (5) =1.34, p > 0.05.

Light colors (M = 7.64, SD = 1.56) were brighter than heavy colors (M =3.68, SD = 2.01), t (21) = 6.06, p < 0.01, but had lower chroma (M = 5.36,SD = 3.51; M = 7.45, SD = 3.50, respectively), t (21) = −2.41, p < 0.05.

Warm colors (M = 9.20, SD = 3.61) had higher chroma than cold colors(M = 5.73, SD = 3.45), t (14) = 2.5, p < 0.05, but did not differ on bright-ness, t (14) = −1.76, p > 0.05.

Multisensory Research (2015) DOI:10.1163/22134808-00002512 15

Thin colors (M = 8.64, SD = 1.03) were brighter than thick colors (M =3.55, SD = 1.69), t (10) = 11.2, p < 0.01, but had less chroma (M = 3.09,SD = 2.43; M = 6.91, SD = 3.51, respectively), t (10) = −3.3, p < 0.05.

Dry colors did not differ from humid colors on brightness t (3) = 0.2, p >

0.05, or chroma, t (3) = −1.26, p > 0.05.For each tactile or haptic term participants could choose an intensity rang-

ing from 1 to 5. The choice of the tactile or haptic adjective was equivalentto the choice of the intensity of the haptic characteristic. Several partial Pear-son’s correlations were run to assess the relationship between brightness andchroma of the target color samples and the tactile and haptic characteristics(smooth, rough, soft, hard, sticky, non-sticky, supple, rigid, elastic, stiff, vis-cous, fluid, light, heavy, warm, cold, thin, thick, pointy, round, dry, humid)assigned to these samples. All correlations were conducted while controllingfor hue. Most of the found correlations with brightness and chroma range frommedium (�0.30) to large (�0.50) effect sizes. The results of this analysis aregiven in Table 1.

3.3. Hue and Tactile and Haptic Adjectives

Term maps and histograms suggest that haptic and tactile terms tend to con-centrate in certain hue ranges. As can be seen in Fig. 3, the ranges of huesmatched to the opposed terms ‘smooth’ and ‘rough’ lie opposite each otheron the hue circle. This seems particularly true in pairs of adjectives one ofwhich has a consensus above 50%. These six pairs are: fluid/viscous, hu-mid/dry, light/heavy, smooth/rough, soft/hard, thin/thick, and warm/cold. InFig. 4, these pairs of opposed adjectives are presented on the Munsell hue cir-cle. The hue for which highest consensus was reached is marked by a circle.These figures reveal that the hues most consensually matched to opposed tac-tile and haptic adjectives do not overlap in color space, with the exception ofthin/thick. Second, in most cases, the opposed adjectives are also opposed incolor space. ‘Soft’ (78% of consensus) is mostly matched to 7.5P, while ‘hard’(69.5%) is matched to 7.5Y. In the Munsell hue circle, 7.5P and 7.5Y are op-posite each other. The same is true for ‘fluid’ (67.8%, 2.5BP) and ‘viscous’(52.5%, 2.5Y), ‘humid’ (71.2%, 2.5PB) and ‘dry’ (47.5%, 7.5Y), ‘smooth’(62.7%, 2.5R), and ‘rough’ (55.9%, 7.5Y). ‘Light’ has two equally consensualpoints (71.2%), one at 2.5YR and the other at 2.5RP, and ‘heavy’ (76.3%) ismatched to 7.5Y. ‘Thin’ also has two most consensual points (49.2%), oneat 2.5BG and the other at 2.5RP, while ‘thick’ (52.5%) is most consensuallymatched to 7.5P. In the case of thin/thick, the reached level of consensus ofwhich is overall lower than in the other six pairs, the hue opposition is notclear.

Across these six pairs, and as can be seen in Fig. 4, one group of adjectives(rough, hard, heavy, dry, viscous) is mostly matched to greenish yellows, while

16 Y. Jraissati et al. / Multisensory Research (2015)

Table 1.Descriptive statistics and correlations

M SD Correlation brightnessa Correlation chromaa

Smooth 2.86 0.55 0.53∗∗ −0.38∗∗Rough 2.69 0.79 −0.2 −0.24∗Soft 2.85 0.56 0.46∗∗ −0.38∗∗Hard 2.83 0.54 −0.32∗ −0.25∗Sticky 2.86 0.52 −0.08 −0.05Non-Sticky 3.14 0.52 0.08 −0.34∗Supple 2.71 0.48 0.4∗∗ −0.3∗Rigid 2.98 0.50 −0.2 −0.35∗∗Elastic 2.79 0.45 0.1 −0.31∗Stiff 2.96 0.72 −0.28∗ −0.05Viscous 2.81 0.65 0.04 −0.09Fluid 2.80 0.52 0.31∗ −0.21Light 2.86 0.71 0.55∗∗ −0.44∗∗Heavy 2.80 0.64 −0.26∗ −0.31∗Warm 2.80 0.64 −0.24 0Cold 2.99 0.58 0.27∗ −0.43∗∗Thin 2.80 0.67 0.39∗∗ −0.31∗Thick 2.58 0.59 −0.42∗∗ −0.21Pointy 2.87 0.57 −0.02 −0.1Round 2.82 0.44 0.34∗ −0.53∗∗Dry 2.89 0.66 −0.32∗ −0.32∗Humid 2.64 0.53 0.36∗ −0.12

a Pearson’s partial correlations (r) controlling for hue.∗ Significant at the p < 0.05 level.

∗∗ Significant at the p < 0.001 level.

the group of opposed adjectives (smooth, soft, light, humid, fluid) is mostlymatched to purples, or to shades close to purples. One of these six pairs of ad-jectives features a different behavior: ‘warm’ (70.7%, 7.5R) and ‘cold’ (78%,2.5B) which are most consensually matched to red and blue respectively.

4. Discussion

In this study, we sought to examine whether (1) participants would match tac-tile and haptic terms to color categories, or (2) match haptic terms to (a) bright-ness, (b) chroma, and to (c) hue, and (3) the proposed experimental paradigmwould clarify the role of hue in cross-modal matching with color.

4.1. Tactile and Haptic Adjectives Are Not Matched to Color Terms

Our results show a consistent pattern of associations between tactile and hap-tic vocabulary pertaining to surface and substance properties and color. In

Multisensory Research (2015) DOI:10.1163/22134808-00002512 17

all cases, except pointy/round, elastic, sticky/non-sticky, supple, and dry, theassociation between haptic and tactile terms and color samples reached or ex-ceeded the threshold of 50% of consensus. This shows that the associationof tactile and haptic adjectives to color is not random. However, as expectedin light of Ludwig and Simner’s (2013) results, we found that the matchingof haptic and tactile terms to color did not fall within the boundaries of colorlexical categories. Colors matched to ‘soft’ were not all called ‘pink’, and con-versely, colors called ‘pink’ were not all matched to ‘soft’.

4.2. Tactile and Haptic Adjectives Form Structured Clusters in Color Space

The distribution of tactile and haptic adjectives in color space suggests thatmany terms that reach or exceed the threshold of 50% consensus in their as-sociation to color feature a radial structure. That is, several tactile and hapticterms form structured clusters in color space, featuring a highly consensualcenter, with levels of consensus decreasing as we move away from the cen-ter of the term’s extension. Such pattern is often reported in the carving ofperceptual and conceptual spaces by a given lexicon. It reflects the degree towhich a given point in space is labeled ‘x’ — or its degree of membership tothe category referred to by the term ‘x’ (Rosch, 1973, 1999; Rosch Heider,1972; Rosch and Mervis, 1975). Radial organization of the extensions of thehaptic adjectives in color space suggests that the similarity relations betweencolors affect the associations. Some color samples are most consistently andconsensually associated to a given tactile or haptic adjective. The less a givencolor patch is similar to the most consensually matched color samples, the lessit is consensually associated to the tactile or haptic adjective. The fact that tac-tile and haptic terms were matched to color in this way also suggests that bothbrightness and hue were relevant to the way participants matched tactile andhaptic terms to color.

4.3. Systematic Association Between Brightness, Chroma, and Tactile andHaptic Adjectives

Our study revealed that color samples associated with ‘smooth’, ‘soft’, ‘sup-ple’, ‘light’, and ‘thin’ were brighter than color samples associated with‘rough’, ‘hard’, ‘rigid’, ‘heavy’, and ‘thick’ respectively. In line with Ludwigand Simner (2013) who showed the role of brightness in associations betweentactile sensations and color, our study confirmed that brightness plays a keyrole in matching tactile and haptic adjectives to color as well.

As for chroma, our study showed that all of ‘smooth’, ‘rough’, ‘soft’, ‘hard’,‘non-sticky’, ‘supple’, ‘rigid’, ‘elastic’, ‘light’, ‘heavy’, ‘cold’, ‘thin’, ‘dry’and ‘round’ were matched to low levels of saturation. This finding is differentfrom the one that could have been expected on the basis of Ludwig and Sim-ner’s (2013) results, where high chroma was matched to softer and smoother

18 Y. Jraissati et al. / Multisensory Research (2015)

Multisensory Research (2015) DOI:10.1163/22134808-00002512 19

stimuli, and low chroma to harder and rougher stimuli. In our study, this dif-ference should be examined in light of the characteristics of the Munsell array,where chroma is high at middle values, in particular in the red-to-green huerange. ‘Rough’, ‘hard’, ‘rigid’, ‘heavy’ and ‘dry’ were matched to dark colorswhere chroma is particularly low. ‘Smooth’, ‘soft’, ‘light’, ‘thin’ were matchedto bright colors where chroma is also low. Finally, ‘cold’ was mostly matchedto colors in the blue–green range, where even at mid-levels of value, chromais average to low. The results of this study ultimately show that specific tactileand haptic terms were matched to particular colors on the basis of the charac-teristics of brightness and chroma, as well as hue.

4.4. Color Opposition in Hues Matched to Tactile and Haptic Adjectives

We observed that there is a tendency to match polar adjectives to non-overlapping ranges of hue, sometimes even opposite hues, in the Munsell huecircle. According to Hering (1964), there are opponency relationships betweenhues that structure our color space. Human experience of color is character-ized by six unique hues, white, black, blue, yellow, red, and green. These huesare unique in comparison to hues such as orange or purple, the appearanceof which can be described in terms of their similarity to yellow and red, andblue and red, respectively. In contrast, red is not described in terms of its sim-ilarity to orange and purple. Red, along with yellow, blue, green, white andblack, can be represented as an ‘unmixed’, ‘pure’ color. Also, while a colorsuch as orange can be more or less reddish or yellowish, there is no such thingas a reddish green, or a yellowish blue. In this sense, blue is opposed to yel-low, and green to red. These relations of opponency yield two orthogonal axes(red/green and yellow/blue) that structure the hue circle and account for sim-ilarity relations between colors. This opposition in color sensation is thoughtto be due to low-level light information processing mechanisms (De Valoiset al., 1966, 2000) and has been shown in behavioral research (Hurvich andJameson, 1955).

In the Munsell system, the orthogonal axes are 10R/5BG (red/green axis)and 10GY/10PB (yellow/blue axis). Note, however, that Romney and Indow(2002) suggested the axes to be 10R/2.5B and 7.5GY/5P. We observed that theopposed terms light/heavy, soft/hard, smooth/rough, humid/dry were opposedin the Munsell array, with ‘light’, ‘soft’, ‘smooth’, and ‘humid’ ranging mostly

Figure 4. Representation of the most consensual hues on the Munsell hue circle for the follow-ing pairs of adjectives, from left to right, top to bottom: smooth/rough, soft/hard, fluid/viscous,humid/dry, warm/cold, light/heavy, and thin/thick. The circles indicate the hue with the highestparticipant consensus for a specific adjective, with each adjective within a pair being representedby a full or empty circle. The dashed lines represent the two orthogonal axes of opposed colorsin Munsell, while the black lines show the relation of the hues matched to opposed adjectives.

20 Y. Jraissati et al. / Multisensory Research (2015)

across the blue–purple hues, and ‘heavy’, ‘hard’, ‘rough’, and ‘dry’ rangingmostly across the yellows. On the other hand, ‘warm’ ranged across the redhues, and ‘cold’ across the blue hues. Thus, most interestingly, light/heavy,soft/hard, smooth/rough, humid/dry were matched to opposite hues along theY/B axis, while warm/cold was matched to opposite hues along the R/G axis,which is orthogonal to the first. These results confirm that hue plays a role incross-modal association between color and haptic and tactile terms.

4.5. Efficiency of the Experimental Paradigm

It should be noted that there are important differences between the currentstudy and the one by Ludwig and Simner (2013) and the comparison betweenthe two should, therefore, be made cautiously. Specifically, as explained in theintroduction, this study did not seek to replicate Ludwig and Simner’s results,but (1) to make sure that the associations are not lexical, as suggested by theirreduction of hue responses to lexical categories, and (2) to focus on the asso-ciation to hue. For this reason, the experimental paradigm was quite different.Accordingly, some differences in the obtained results can be attributed to thepeculiarities of the used paradigms. For example, the use of the Munsell ar-ray, as opposed to the RGB color wheel used by Ludwig and Simner, mostcertainly explains the different association patterns to saturation.

The difference in the direction of the matching task should also be noted.In our study, participants saw individual color samples and had to match tac-tile and haptic adjectives to these individual colors, rather than match tactilesensations to colors. The rationale for this inversion was to avoid the inter-ference of memory for previously made associations. If the participants werepresented with haptic/tactile terms first (e.g., ‘soft’) and then asked to asso-ciate them to a color on the Munsell array, chances that they would associatean opposed term (e.g., ‘hard’) to the same or neighboring colors would havebeen low due to the memory of the previously made associations. On the otherhand, the memory bias is less likely if the colors are presented one at a timeand the tactile and haptic terms are selected to match the colors.

Finally, an obvious difference is that in this study we used haptic and tactileterms, rather than sensations. It is possible that the use of terms might havefavored the regularity in the associations. But this is not a question we soughtto address in this study. The use of the terms was necessary to rule out thepossibility of a lexical nature of the associations. Additionally, this paradigmenabled addressing the role of hue, which might not have been investigatedadequately in Ludwig and Simner’s (2013) study given their method of re-ducing participants’ selection of hue on the color wheel to lexical categories.While this method might indicate whether tactile sensations are matched tolexical categories, it is not informative of whether or not tactile sensations arematched to hue. For example, Ludwig and Simner conclude that ‘soft’ is not

Multisensory Research (2015) DOI:10.1163/22134808-00002512 21

matched to ‘pink’, because ‘soft’ was only matched to bright shades of ‘pink’.However, the lexical category ‘pink’ in English typically ranges across severalhues, spanning from purple to red, in the Munsell array at medium to high lev-els of brightness (Berlin and Kay, 1969; Davidoff et al., 1999). Therefore, thematch between ‘soft’ and bright ‘pink’ found by Ludwig and Simner, could,in fact, be the match between ‘soft’ and the purple–red hue range, which isconsistent with our current study. They also noted that ‘rough’ was matched todark ‘brown’, which suggests that ‘rough’ was probably matched to the yellowhue range, at low levels of brightness. This is also in line with our findings.However, because of the reduction of hue responses to lexical categories, suchassociations could have been missed in Ludwig and Simner’s data. Therefore,we consider the paradigm developed in this study to be successful to the extentthat the Munsell array allows a clear examination of the association of hapticand tactile terms to hue (also see the remarks on hue in Slobodenyuk et al.,2015).

4.6. Touch, Haptics and Color: Possible Accounts of the ObservedSystematic Associations

The fact that in many instances opposed haptic terms were matched to non-overlapping, sometimes even opposite, colors in Munsell was unexpected. Thereason why such patterns were observed is open to question.

It is possible that the associations between haptic/tactile terms and colorsensations are conceptual. However, in this study, we only addressed, and over-ruled, the possibility of the associations being lexical. It would be interestingto assess the role of concepts and representations in association of haptic andtactile property terms to color in future research.

Another possible explanation of the color and tactile associations pertainsto the regularities in the environment (Parise and Spence, 2009). We observed,for example, that ‘fluid’ was mostly matched to blue. It is possible in this casethat the color blue was matched to ‘fluid’ because fluids tend to be bluish in ourenvironment. However, dark brownish colors were also matched to ‘rough’,‘hard’, and ‘heavy’, yet not all things brown in our environment are rough,hard, or heavy. Mud, for example, is soft, not hard, even if relatively heavy(Ludwig and Simner, 2013).

Apart from color, perceptual experience in different modalities is usu-ally represented by structured spaces, featuring axes linking opposing poles(Levinson and Majid, 2014; Murphy, 2011). These spaces represent relationsof similarity and difference between stimuli, where a geometrical distance inthe space represents the perceptual distance between the stimuli. Studies ofperceptual and semantic-perceptual space of touch are abundant (BergmannTiest et al., 2006; Hollins et al., 1993, 2000; Picard et al., 2003; Soufflet et al.,2004; Yoshida, 1968). Across these studies, some dimensions of tactile expe-

22 Y. Jraissati et al. / Multisensory Research (2015)

rience seem recurrent, although some discrepancies can be noted as well (pos-sibly due to native language of participants, preferred number of dimensionsin multi-dimensional scaling, pleasantness, and stimulus materials [Guest etal., 2011]). Apart from soft/hard, smooth/rough, which are considered to berobust dimensions (Hollins et al., 2000; Picard et al., 2003), sticky/slipperyhas also been put forward (Hollins et al., 1993). However, in other studiessticky/slippery and warm/cold dimensions were not found to be independentof soft/hard and smooth/rough (Picard et al., 2003), while thin/thick seemsto require haptic manipulation of the stimuli (Soufflet et al., 2004). Guest etal. (2011), who argue that they offer the most global or archetypical space oftactile experience, suggest that smooth/rough, hot/cold, and dry/wet are inde-pendent dimensions of tactile experience. Soft/hard did not seem independentof smooth/rough in their study.

In our study, smooth/rough, soft/hard, light/heavy, humid/dry, fluid/viscous,thin/thick and warm/cold were pairs of terms at least one of which reached adegree of consensus greater than 50%. All these terms, except thin/thick, aremore or less matched to the orthogonal color axes. The fact that most of thesetactile terms, which have been suggested to name the poles structuring tactilespace, are matched to poles in color space cannot be trivial. Levinson and Ma-jid (2014) note that some dimensions of perceptual experience are encoded,or named, in the same way across modalities: sounds and lights are bright ordull; noises and objects are big; sounds and colors are loud or soft; surfacesand pains are hard, etc. They argue that such dimensions being named in thesame away across modalities could be significant with regard to how thesemodalities are taken to relate to each other. However, as Guest et al. (2011)remark, the relationship between the dimensions structuring a given space andthe lexicon used to describe them is not clear: “While lexical descriptors nec-essarily map into a perceptual space, it is unclear what specific descriptorsare consistently used to describe the underlying dimensions of touch” (p. 532;also see Malt et al., in press, for a similar suggestion). In the case of our study,although hue dimensions cannot be said to carry the same names as tactiledimensions (as in the cases of bigness, loudness, intensity, brightness, dull-ness, etc.), there might nevertheless be an instance of matching of dimensionsbetween color (including hue) and touch.

Finally, Palmer et al. (2013) suggested that emotions might mediate cross-modal associations. With regard to the color and music stimuli used in theirstudy, it was proposed that people would pick colors based on their emotionalresponse to music they listen to. Palmer et al. (2013) observed that faster mu-sic in major mode was matched to colors that were more saturated, lighter andin the yellow hue range, while slower music in minor mode was matched todesaturated, darker colors and in the blue hue range. Emotional ratings of col-ors and music were obtained separately. A multi-dimensional scaling of these

Multisensory Research (2015) DOI:10.1163/22134808-00002512 23

emotional responses in both color and music produced 2D spaces with the fol-lowing two dimensions: positive/negative valence (happy/sad) and high/lowpotency (strong/weak). In a separate experiment in the same study, Palmer etal. (2013) also observed that colors were matched to neutral/calm, sad, happy,and angry faces following patterns similar to the ones observed in the caseof music. The authors take these results to strongly support the view that atleast some cross-modal correspondences, namely between music and color,are based on emotion.

In light of these results, the possibility that emotions might be mediating theassociation of color to touch cannot be ignored. In their study of the semantic-perceptual space of tactile experience, Guest et al. (2011) indicate that it isstill unclear how many emotional dimensions exist in this perceptual spaceand if they are related to its established sensory dimensions. Based on theirstudy of tactile lexicon, which includes words referring both to the purelysensory and emotional aspects of tactile experience, the authors obtained a3D semantic-perceptual space using a multi-dimensional scaling technique,with the following dimensions: smooth/rough, hot/cold, dry/wet. They nextobtained ratings of the adequacy of these sensory and emotional words in re-sponse to textured materials moved on several parts of the upper body. Theseratings revealed that greater comfort was associated with reducing roughness,and greater arousal with increasing roughness, firmness and pile. Thus, someemotional dimensions do seem to be related to established dimensions of tac-tile experience (Guest et al., 2011). It is, therefore, possible that in the caseof the touch/color interface, as in the case of the music/color interface, theassociations might be mediated by emotions.

Acknowledgements

Data collection for this study was done with the help of Reem Hmaidan andSara Michli.

References

Bergmann Tiest, W. M. and Kappers, A. M. I. (2006). Analysis of haptic perception of materialsby multidimensional scaling and physical measurements of roughness and compressibility,Acta Psychol. 121, 1–20.

Berlin, B. and Kay, P. (1969). Basic Color Terms: Their Universality and Evolution. CLSIPublications, Stanford, CA, USA.

Bremner, A., Caparos, S., Davidoff, J., de Fockert, J., Linnell, K. and Spence, C. (2013). Boubaand Kiki in Namibia? A remote culture make similar shape-sound matches, but differentshape–taste matches to Westerners, Cognition 126, 165–172.

24 Y. Jraissati et al. / Multisensory Research (2015)

Cook, R., Kay, P. and Regier, T. (2005). The World Color Survey database: history and use,in: Handbook of Categorization in the Cognitive Sciences, H. Cohen and C. Lefebvre (Eds),pp. 224–242. Elsevier, Amsterdam, The Netherlands.

Crisinel, A.-S. and Spence, C. (2010). As bitter as a trombone: synesthetic correspondences innonsynesthetes between tastes/flavors and musical notes, Attent. Percept. Psychophys. 72,1994–2002.

Crisinel, A.-S., Cosser, S., King, S., Jones, R., Petrie, J. and Spence, C. (2012). A bittersweetsymphony: systematically modulating the taste of food by changing the sonic properties ofthe soundtrack playing in the background, Food Qual. Pref. 24, 201–204.

Davidoff, J., Davies, I. and Roberson, D. (1999). Colour categories in a stone-age tribe, Nature398, 203–204.

De Valois, R. L., Abramov, I. and Jacobs, G. H. (1966). Analysis of response patterns of LGNcells, J. Opt. Soc. Am. 56, 966–977.

De Valois, R. L., De Valois, K. K. and Mahon, L. E. (2000). Contribution of S opponent cellsto color appearance, Proc. Natl Acad. Sci. 97, 512–517.

Deroy, O. and Spence, C. (in press). Crossmodal correspondences: four challenges, Multisens.Res. DOI:10.1163/22134808-00002488.

Deroy, O., Crisinel, A.-S. and Spence, C. (2013). Crossmodal correspondences between odorsand contingent features: odors, musical notes, and geometrical shapes, Psychonom. Bull.Rev. 20, 878–896.

Gallace, A. and Spence, C. (2006). Multisensory synesthetic interactions in the speeded classi-fication of visual size, Percept. Psychophys. 68, 1191–1203.

Google books Ngram Viewer (n.d.). Retrieved from https://books.google.com/ngrams/.Guest, S., Dessirier, J.-M., Mehrabyan, A., McGlone, F., Essick, G., Geschedier, G., Fontana,

A., Xiong, R., Ackerley, R. and Blot, K. (2011). Attent. Percept. Psychophys. 75, 531–550.Hering, E. (1964). Outlines of a Theory of the Light Sense. Harvard Univeristy Press, Cam-

bridge, MA, USA.Ho, H. N., Van Doorn, G. H., Kawabe, T., Watanabe, J. and Spence, C. (2014). Colour–

temperature correspondences: when reactions to thermal stimuli are influenced by colour,PLoS One 9, e91854. DOI:10.1371/journal.pone.0091854.

Hollins, M., Faldowski, R., Rao, S. and Young, F. (1993). Perceptual dimensions of tactilesurface texture: a multidimensional scaling analysis, Percept. Psychophys. 54, 697–705.

Hollins, M., Bensmaïa, S., Karlof, K. and Young, F. (2000). Individual differences in perceptualspace for tactile textures: evidence from multidimensional scaling, Percept. Psychophys. 62,1534–1544.

Hurvich, L. M. and Jameson, D. (1955). Some quantitative aspects of an opponent-colors theory.I. Chromatic responses and spectral saturation, J. Opt. Soc. Am. 4, 546–552.

Jraissati, Y. (2010). Purple in Lebanese Lexicons. Unpublished manuscript.Karwoski, T. F., Odbert, H. S. and Osgood, C. E. (1942). Studies in synesthetic thinking: II. The

role of form in visual response to music, J. Gen. Psychol. 26, 199–222.Kim, Y.-J. (2013). Can eyes smell? Cross-modal correspondences between color hue-tone and

fragrance family, Color Res. Appl. 38, 139–156.Köhler, W. (1929). Gestalt Psychology. Liveright, New York, NY, USA.Levinson, S. C. and Majid, A. (2014). Differential ineffability and the senses, Mind Lang. 29,

407–427.

Multisensory Research (2015) DOI:10.1163/22134808-00002512 25

Li, P., Zhang, F., Tsai, E. and Puls, B. (2014). Language history questionnaire (LHQ 2.0): a newdynamic web-based research tool, Biling. (Camb. Engl.) 17, 673–680.

Ludwig, V. U. and Simner, J. (2013). What colour does that feel? Tactile–visual mapping andthe development of cross-modality, Cortex 49, 1089–1099.

Malt, B., Gennari, S., Immai, M., Ameel, E., Saji, N. and Majid, A. (in press). Where are theconcepts? What words can and can’t reveal, in: Concepts: New Directions, E. Margolis andS. Laurence (Eds). MIT Press, Cambridge, MA, USA.

Marks, L. E. (1974). On associations of light and sound: the mediation of brightness, pitch andloudness, Am. J. Psychol. 87, 173–188.

Martino, G. and Marks, L. E. (1999). Cross-modal interaction in vision and touch, poster pre-sented at the 40th Annual Meeting of the Psychonomics Society, Los Angeles, CA, USA.

Melara, R. and Marks, L. E. (1990). Processes underlying dimensional interactions: correspon-dences between linguistic and nonlinguistic dimensions, Mem. Cogn. 18, 477495.

Monroe, M. (1925). The apparent weight of color and correlated phenomena, Am. J. Psychol.36, 196–206.

Munsell, A. H. (1941). A Color Notation: An Illustrated System Defining All Colors and TheirRelations. The Hoffman Borthers Co., Boston, MA, USA.

Murphy, V. N. (2011). Olfactory maps in the brain, Annu. Rev. Neurosci. 34, 233–258.Newhall, S., Nickerson, D. and Judd, D. (1943). Final report of the OSA sub-committee on the

spacing of the Munsell colors, J. Opt. Soc. Am. 33, 385–418.Palmer, S. E., Schloss, K. B., Xu, Z. and Prado-León, L. R. (2013). Music–color associations

are mediated by emotion, Proc. Natl Acad. Sci. 110, 8836–8841.Parise, V. C. and Spence, C. (2009). ‘When birds of a feather flock together’: synesthetic

correspondences modulate audiovisual integration in non-synesthetes, PLoS One 4, e5664.DOI:10.1371/journal.pone.0005664.

Picard, D., Dacremont, C., Valentin, D. and Giboreau, A. (2003). Perceptual dimensions oftactile textures, Acta Psychol. 114, 165–184.

Ramachandran, V. S. and Hubbard, E. M. (2001). Synaesthesia — a window into perception,thought and language, J. Consc. Stud. 8, 3–34.

Romney, A. K. and Indow, T. (2002). A model for the simultaneous analysis of reflectance spec-tra and basis factors of Munsell color samples under D65 illumination in three-dimensionalEuclidean space, Proc. Natl Acad. Sci. 99, 11543–11546.

Rosch, E. (1973). Natural categories, Cogn. Psychol. 4, 328–350.Rosch, E. (1999). Principles of categorization, in: Concepts: Core Readings, E. Margolis and

S. Laurence (Eds), pp. 189–206. MIT Press, Cambridge, MA, USA.Rosch, E. and Mervis, C. B. (1975). Family resemblances: studies in the internal structure of

categories, Cogn. Psychol. 7, 573–605.Rosch Heider, E. (1972). Probabilities, sampling, and ethnoraphic method: the case of Dani

colour names, Man 7, 448–466.Schifferstein, H. N. J. and Tanudjaja, I. (2004). Visualising fragrances through colours: the

mediating role of emotions, Perception 33, 1249–1266.Slobodenyuk, N., Jraissati, Y., Kanso, A., Ghanem, L. and Elhajj, I. (2015). Cross-modal asso-

ciations between color and haptics, Attent. Percept. Psychophys. 77, 1379–1395.Soufflet, I., Calonnier, M. and Dacremont, C. (2004). A comparison between industrial experts’

and novices’ haptic perceptual organization: a tool to identify descriptors of the handle offabrics, Food Qual. Pref. 15, 689–699.

26 Y. Jraissati et al. / Multisensory Research (2015)

Spence, C. (2011). Crossmodal correspondences: a tutorial review, Attent. Percept. Psychophys.73, 971–995.

Tomasik-Krótki, J. and Strojny, J. (2008). Scaling of sensory impressions, J. Sens. Stud. 23,251–266.

Walker, P. (2012). Cross-sensory correspondences and cross talk between dimensions of con-notative meaning: visual angularity is hard, high-pitched, and bright, Attent. Percept. Psy-chophys. 74, 1792–1809.

Walker, P., Francis, B. and Walker, L. (2010). The brightness-weight illusion: darker objectslook heavier but feel lighter, Exp. Psychol. 57, 462–469.

Yoshida, M. (1968). Dimensions of tactual impressions, Jpn. Psychol. Res. 10, 157–173.