$1."~‘i.5‘\"{: '91“! ‘ “'1‘“ \ {1‘94 9 ‘nflr’? “,9, Ii "3va . i '> ",1 , a J-‘mui 1)., 95:23:??? ‘ M36 r w. .3 .M. V . ’ (:5 -4, \- -. . an .‘erf'u r o . 3.3," ,.- . .‘ mm a-.. ,0. . n. u- «71”.... , ,f....'. ,. .. w .. 1 21’ ‘ . 5334::- n a; UN; , v, “warms- '4'- ,,—".’2'.2‘..'."a‘.737““ ‘9 (7 ' 4' J. ... o . .. . no-rrr » 44-19. .- NP“ T I, t 1’ 5 Lawn“, ‘ m. ‘ ,y'a , . f“ . j 3; .\I.¢‘.‘¢r'1£ -: 33—5- lrrrt-lr _ a ..1. . A .mw - , ..., may . a . :u, -=i.Ivu-'!!\- , yr. - v lfi‘ ' .5. "'50: 2?. Ann- I,..U;---"L., ‘ .CLJ Y N . ._. , a J ,, ‘ ,., .. "41'5'7'. -”<-'."!5" f""‘*.:_- ‘3'. r1 r- 1. . :r'..._:fl....-!..., .. -- * u '2 .... . n4 ”r... ... . w. .- -,. .14.“? ... ‘J‘..-.'.‘Ju 'r' 1’ .r M17,- .. s’or’» . .--......'..'..'$'l Ml iCHl IGAN STATEU I I IIII III IIIIIIIIIIIIIII IIIIIIIIII 3 129 0096 4573 II This is to certify that the dissertation entitled THE PREDICTION OF SURROUND-INDUCED CHANGES IN MAP COLOR APPEARANCE presented by Cynthia Ann Brewer has been accepted towards fulfillment of the requirements for Ph . D . degree in Geography I Major professor Date 7/24/91 MS U is an Affirmative Action/Equal Opportunity Institution 0- 12771 I" —\ ‘ LIBRARY Michigan State University I \_ __ \ A PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. , DATE DUE DATE DUE DATE DUE my 1 U ”TIL {avail “'1 “6‘0 fig A I n—u fl : __I MSU Is An Affirmative Action/Equal Opportunity Institution emu-drama” * THE PREDICTION OF SURROUND-INDUCED CHANGES IN MAP COLOR APPEARANCE By Cynthia Ann Brewer \a. A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Geography 1991 g 54—. 5 6795 ABSTRACT THE PREDICTION OF SURROUND-INDUCED CHANGES IN MAP COLOR APPEARANCE By Cynthia Ann Brewer The overall objective of my research was to develop a quantitative model of simultaneous contrast (induction) to aid selection of sets of easily identified map colors. Modelling induction allows use of the smaller color differences needed to design perceptually logical color schemes while avoiding surround- induced misinterpretations of mapped information. Developing objective guidance for color selection is important because most people who work with geographic data in a digital environment are confronted with the problem of choosing colors for the visual display and analysis of their data. The research addressed two primary questions: does a color shift in the direction of the opponent complement of its surround, as suggested by review of previous literature, and what is the magnitude of the induced shift? A first experiment, with test colors embedded in hexagonal grids, was used to quantify the distances and directions in color space of changes in appearance. The resulting model of induction is an extension of the color appearance model developed by R.W.G. Hunt. Predictions from the induction model were tested in a second map-reading experiment. The induction model can be used to provide an objective evaluation of a set of map colors. To predict lightness induction, the contrast in lightness between a center color and its surround is calculated. A small proportion of this contrast is then added to the color of the center to estimate its perceived lightness. Shifts in hue and saturation are estimated by constructing a buffer, Cynthia Ann Brewer or zone of potential perceptions, that extends from the color toward the opponent complement of the surrounding hue. If the center and surround are similar in hue, the buffer extends a greater distance than if the center and surround are dissimilar in hue. If induction causes colors to have overlapping hue and lightness perception buffers, these colors will be difficult to distinguish on the map. The buffer approach was chosen to accommodate 90 percent of map readers’ perceptions, rather than limiting conclusions to perceptions of an artificial average reader. The model developed provides a practical tool for the design of successful color schemes for computer- displayed maps. C0pyright by Cynthia Ann Brewer 1991 ACKNOWLEDGEMENTS Judy Olson’s role as my advisor is acknowledged with appreciation. My year as an assistant for her NSF research provided an important model of why outside funding is important and how to conduct a full-scale research project. The pilot study for the dissertation research was a supporting investigation funded by her grant. Her sabbatical and appointment as department chair provided the positive intangible of pushing me toward becoming an independent researcher. Iudy provided a year’ 5 worth of space and equipment that made the project possible: her computer, a 19-inch monitor, and her office (which I monopolized with a weird-looking black booth). Advising from my committee members is appreciated. Dr. Winters kept me focussed on my roots in geography as I voyaged into color science. His advice on planning my future (on types of teaching positions, on service, on publication goals, on money) gave me a realistic view of my profession. Dr. Irwin’s expertise in writing convincing grants was crucial to the success of my funding proposal to NSF. He also provided important analysis recommendations. From the beginning, Dr. Groop offered encouragement that this topic was important. ' The people of the Inter-Society Color Council were an important source of information for me as I tackled color theory. The efforts of the ISCC Vision and Color Appearance Interest Group are appreciated; they brought RW.G. Hunt to the US. for the 1989 meeting. Dr. Hunt’ 5 careful attention to what I V was planning to do with his model was valuable. After that meeting, I was confident of the soundness of my proposal. This research was supported by National Science Foundation Dissertation Research Improvement Grant Number 5135-8915175. The grant was used primarily to purchase a color-measurement instrument, pay test subjects, and pay the wages of three undergraduate assistants. The funding was crucial to the dissertation research. The three students who assisted the research kept me sane by helping with the monotonous task of subject testing. Mike Conley and Carolyn Lauer also assisted in the laborious process of color assignments for Experiment 1. Eva Frank designed a set of beautiful color schemes for the second experiment. All three assistants engaged me in discussions that lead to refinements of the research in progress. For clarification of objectives, there is no substitute for instructing another person in how to do what you want accomplished. In the main office of the Department of Geography, Harriet Ashbay organized the scheduling of my experiment volunteers and Marilyn Bria helped me with the mysteries of grant logistics (i.e., how to spend that NSF money). David DiBiase solved many creative and technical puzzles with me. He endured many hours of the explanation of half-formed ideas, and he was always able to pinpoint the crux of a problem and generate unexpected solutions. He helped with the nitty-gritty of what software to use, how to move graphics between cantankerous programs, how to cure ailing Macintoshes, and generally how to get things to work. He likened this adventure to standing at the mouth of a deep cave shouting directions down to me so I did not get lost. vi TABLE OF CONTENTS LIST OF TABLES .......................................................................................................... xi LIST OF FIGURES ........................................................................................................ xiii INTRODUCTION ........................................................................................................ 1 CHAPTER 1 Review of Terminology and Literature .................................................................. 4 Terminology ..................................................................................................... 5 Cartographic Literature ................................................................................... 7 Simultaneous Contrast Literature ............................................................... 11 Brightness and Lightness Induction ...................................................... 11 Hue Induction and Complementarity .................................................. 16 Saturation or Chroma Induction ........................................................... 20 Test Methods ............................................................................................... 21 Spatial Aspects of Induction .................................................................... 22 Quantitative Generalization about Induction ........................................... 25 Modelling Color Appearance .................................................................. 27 Summary ........................................................................................................... 31 vii CHAPTER 2 Experiment 1: Methods .............................................................................................. 33 Test Procedure .................................................................................................. 33 Subjects ......................................................................................................... 34 Test Displays ...................................................................................................... 35 Test Colors ......................................................................................................... 45 Production of Test Colors ......................................................................... 49 Limitations on Color Precision ............................................................... 50 CHAPTER 3 Experiment 1: Results and Discussion ................................................................... 56 Size ...................................................................................................................... 58 Controls .............................................................................................................. 60 Lightness Test ................................................................................................... 64 Hue Distance Test ............................................................................................ 69 Hue Direction Test ........................................................................................... 73 Light and Dark Hue Distance Test ................................................................ 77 Variations on Displays .................................................................................... 80 Hexagon Outlines ...................................................................................... 81 Heterogeneous Surrounds ....................................................................... 83 Summary ........................................................................................................... 85 viii CHAPTER 4 Experiment 2: Methods, Results, and Discussion ................................................ 87 Methods ............................................................................................................. 87 Subjects ......................................................................................................... 87 Test Displays ................................................................................................ 88 Test Procedure ............................................................................................ 94 Hypotheses ........................................................................................................ 97 Results ................................................................................................................ 98 Response Time Data Processing ............................................................. 98 Fulfillment of ANOVA Assumptions .................................................. 101 ANOVA Results ........................................................................................ 102 Graphic Data Summary ............................................................................ 103 Discussion .......................................................................................................... 107 Faulty Adjustments to Schemes ............................................................. 107 Desaturated Surround Effects .................................................................. 108 Summary ........................................................................................................... 111 CHAPTER 5 Conclusions and Recommendations ...................................................................... 112 Lightness Induction ......................................................................................... 112 Chromaticity Induction .................................................................................. 113 Saturation Importance .............................................................................. 113 Recommendations for Saturation Research ........................................ 114 Model Applications ......................................................................................... 115 Recommendations for Applications Research .................................... 116 Final Comments .............................................................................................. 117 ix LIST OF REFERENCES ............................................................................................... 119 APPENDD( A Experiment Procedures ............................................................................................... 125 APPENDIX B Experiment 1 Color Information .............................................................................. 136 APPENDIX C Experiment 1 Frequency Response Data ................................................................. 141 APPENDIX D Program Code for Hunt Model of Color Appearance .......................................... 149 APPENDIX E Experiment 2 Color Information .............................................................................. 175 Table 1.1 Table 1.2 Table 3.1 Table 3.2 Table 3.3 Table 3.4 Table 3.5 Table 3.6 Table 3.7 LIST OF TABLES Perceptual and Related Psychophysical Color Terminology ...................................................................................... 7 Rank Discrepancies in Muller's Continuous-Tone Results ................................................................................................. 10 Somers's dyx Comparison of Change Induced in Small and Large Test Centers ......................................................... 59 dyx Comparison of Change Induced by Control and Inducing Surrounds ................................................................. 63 Percentage of Matches to Comparison Choices for Lightness Distance Test .................................................................... 66 Percent of Matches to Comparison Choices for Hue Distance Test ............................................................................. 72 Frequencies of Matches to Comparison Choices for Hue Direction Test ............................................................................ 75 Percent of Matches to Comparison Choices for Saturated Hue Distance Test ........................................................... 79 dyx Comparison of Change Induced by Outlined Surrounds Relative to Contiguous and Control Surrounds ........................................................................................... 82 xi Table 3.8 Table 4.1 Table 4.2 Table 4.3 Table 8.1 Table C.1 Table C.2 Table C.3 Table C.4 Table E.1 dyx Comparison of Change Induced by Heterogeneous Surrounds Relative to Homogeneous and Control Surrounds ........................................................................................... 84 Experiment 2 Map Titles and Scheme Types .............................. 90 Experiment 2 Maps and Number of Incorrect Responses ........ 99 Experiment 2 AN OVA of Logged Response Times .................. 100 Experiment 1 Color Specifications ................................................ 137 Frequency Response Data for Lightness Distance Test ............. 142 Frequency Response Data for Hue Distance Test ....................... 145 Frequency Response Data for Light and Dark Hue Distance Test ...................................................................................... 147 Frequency Response Data for Hue Direction Test ..................... 148 Experiment 2 Color Specifications ................................................ 176 xii Figure 1.1 Figure 1.2 Figure 1.3 Figure 2.1 Figure 2.2 Figure 2.3 Figure 2.4 Figure 2.5 Figure 2.6 Figure 2.7 Figure 2.8 Figure 2.9 Figure 2.10 Figure 2.11 LIST OF FIGURES Gilchrist’ 5 Experiment Investigating Lightness Constancy and Contrast ................................................................... 14 Comparison of Complementary Color Definitions .................. 18 Hunt Opponent Axes ....................................................................... 30 Experiment 1 Test Display Format ................................................ 37 Test Display of Medium-Sized Hexagons .................................... 38 Test Display of Small-Sized Hexagons ......................................... 39 Test Display with Black-Outlined Hexagons ............................... 41 Test Display with White-Outlined Hexagons ............................. 42 Test Display with Example Heterogeneous Inducing Surround ........................................................................... 43 Experiment 1 Sampling of Hunt Color Space ............................. 48 Center Measurement Variations for Medium-Lightness Hues ..................................................................................................... 51 Center Measurement Variations for Light and Dark Saturated Hues .................................................................................. 52 Measurement Variations for Center Lightness .......................... 53 Experiment 1 Control Display Format ......................................... 55 xiii Figure 3.1 Figure 3.2 Figure 3.3 Figure 3.4 Figure 3.5 Figure 3.6 Figure 3.7 Figure 4.1 Figure 4.2 Figure 4.3 Figure 4.4 Figure 4.5 Figure D.1 Figure D.2 Figure D.3 Figure D.4 Figure D.5 Figure D.6 Comparison-Color Choices Plotted in Hunt Color Space for Control Surround Test .............................................................. 62 Scatterplot for Mean-Response Lightness Data .......................... 67 lightness Induction Prediction Buffers ....................................... 68 Scatterplots of Mean-Response Hue Data .................................... 71 Comparison-Color Choices Plotted in Hunt Color Space for Hue Distance Test ....................................................................... 72 Comparison-Color Choices Plotted in Hunt Color Space for Hue Direction Test ..................................................................... 76 Comparison-Color Choices Plotted in Hunt Color Space for Saturated-Hue Distance Test .................................................... 79 Black-and-White Versions of Example Test Maps .................... 89 Example Color-Pair Selections from Experiment 2 Map Schemes ..................................................................................... 92 Example Arrow Display and Corresponding Map ..................... 96 Graphic Summary of Experiment 2 Results ............................... 105 Comparison of Desaturated Surround Predictions for Maps 1 and 8 ....................................................................................... 110 Card 1 ................................................................................................... 150 Card 2 ................................................................................................... 156 Card 3 ................................................................................................... 159 Card 4 ................................................................................................... 161 Card 5 ................................................................................................... 165 Card 6 ................................................................................................... 170 xiv INTRODUCTION Color selection is a critical aspect of visual display and visual analysis of geographic information. Maps concisely summarize large spatial data sets and are produced for wide-ranging purposes: choropleth maps help us understand distributions of socio-economic characteristics; image maps are one component in the evaluation of remotely-sensed data analyses such as land-cover classifications; and geographic information system maps are an avenue by which we can synthesize the relationships between data layers within complex data sets. In each of these contexts, an easily interpreted color scheme may aid in solving problems or gaining new insights from spatial data. A poor scheme may mask relationships and dilute efforts to convince colleagues of the importance of the work. As cartographers we strive to choose easily distinguished colors that allow accurate recognition of map symbols. We also try to design color schemes that will communicate the logical relationships between mapped data categories or features so that the general form of the mapped distribution is readily apparent. Colors of a carequy designed scheme, however, can be rendered unexpectedly ambiguous by changes in appearance induced by contrast with surrounding map colors. The overall objective of my research is to develop a quantitative model of induction to aid selection of sets of easily identified map colors. Modelling induction allows use of the smaller color differences needed to design perceptually logical color combinations while avoiding surround-induced 1 2 misinterpretations. Understanding color interaction allows map schemes to be designed with confidence. Developing objective guidance for color selection is important because color is no longer an expensive luxury for the computer-graphics tools geographers use. Color display is now unavoidable, and most people who work with geographic data in a digital environment grapple with the problem of choosing colors for the graphic display of their work. The research reported in this dissertation was designed to address the following questions for the entire range of colors on graphic displays: Do colors shift in the direction of the opponent complement of the surround, as suggested by review of previous literature? Are smaller color areas shifted in appearance by greater amounts? What are the magnitudes of induced shifts in a map-reading context? Two secondary questions were also examined in the first experiment: What is the effect of breaking the inducing surround into a heterogeneous combination of colors? How do polygon outlines affect the magnitude of induced changes in appearance? The above questions were shaped into specific hypotheses that are listed along with the results tables in Chapter 3. The results of the first experiment were used to develop a model for prediction of induced changes in the appearance of map colors. A second experiment tested these predictions in a map-reading context. This dissertation is structured into six major divisions. I begin with a chapter reviewing color terminology and the simultaneous contrast literature. The first experiment is reported in Chapters 2 and 3 and the second is reported in Chapter 4. The concluding chapter is followed by references and 3 a set of five appendices. These appendices provide listings of experimental procedures, test-color specifications, response data, and the program code for the model of color appearance upon which my research is based. CHAPTERl Review of Terminology and Literature Simultaneous contrast or surround-induced change has long been recognized as affecting the perception of map colors, and I will review these discussions from the cartographic literature. Little research on simultaneous contrast has been conducted with map stimuli, but both applied and theory- oriented work on induction by color-science, psychology, physiology, photography, and engineering researchers has been conducted. The collection of literature from diverse disciplines, dated over a forty-year time span, reveals a startling array of color terms used with varying precision. Thus, this review chapter begins with definitions and comparisons of color terms. The terminology issues resurface in the review of simultaneous contrast literature because differences in the use of terms complicate the comparison of results and their application within map reading contexts. Past attempts to model induction effects are reviewed with the intent of identifying a fruitful modelling approach on which to build. The first step in predicting simultaneous contrast is the use of numbers to describe perceived colors, and numerous color specification systems exist from which to choose a promising system. Thus, I complete this chapter by discussing the selection of an appearance—based color metric for quantifying color perception. Terminology Generally, simultaneous contrast is the induced enhancement of differences between a given color and surrounding colors; it is the effect of the surround on the color one perceives. Contrast effects are commonly defined with a standard demonstration: a medium gray with a black surround looks lighter than the same gray with a white surround. Peculiar definitions, promulgated through the cartographic literature, designate ’induction’ only for changes immediately along edges (Mach bands) or for brightness contrast effects but not for chromatic effects. The definitions also suggest ’simultaneous contrast’ should be used only for chromatic effects (Castner 1980 p. 374, Dent 1990 p. 385, Peterson 1979 p. 30). These distinctions may have arisen from Robinson’s specific uses of the terms early in the cartographic color literature (Robinson 1952 pp. 93-94, Robinson 1967 p. 54). In current use, however, it is best to avoid this over-specificity that is not common to other disciplines. I will use the terms ’simultaneous contrast’ and ’induction’ as synonyms (largely because ’induction,’ ’induced,’ ’inducing,’ are more flexible for communication). Induction is a general term that encompasses other effects that alter perceptions such as assimilation, by which adjacent small color areas appear more similar (Fach and Sharpe 1986). However, almost every combination of the terms ’simultaneous,’ ’induction,’ and ’contrast,’ appear in the research literature to label simultaneous contrast effects: chromatic induction, brightness induction, simultaneous color induction, simultaneous brightness induction, simultaneous contrast, simultaneous color contrast, chromatic contrast, and brightness contrast. 6 As one tries to make sense of the wide range in terminology that appears in communications about color, one must evaluate the rigor with which terms are used. Terms used as synonyms in lay contexts, such as lightness and brightness, have specific differences in meaning in scientific contexts, and these differences may not be clarified by (or even recognized by) authors. Hue, brightness, and saturation are generally an adequate set of terms to describe perceived color, though there are many more choices in terminology available for color description. Both Hunt (1987a pp. 69-72) and Agoston (1987 p. 12-14) provide good summaries of subjective color terminology, and differences between their definitions are indicative of the lingering lack of consistency among color scientists. Perceptual terms describe subjective color appearance which is affected by the characteristics of both the observer and the object. Psychophysical color specifications are unaffected by changes in the observer. The objective terms of psychophysics are based on color measurements coupled with standardized observer characteristics. Perceptual color terms and their psychophysical correlates are listed in Table 1.1. Hue, brightness, lightness (or value), chroma, and saturation are common perceptual color terms. Hue is the dimension described by color names such as blue and orange. Brightness is dependent either on illumination, for reflecting surfaces, or on emitted light and describes the amount of light that a color exhibits (ranging from dim to dazzling). Lightness is relative brightness; the brightness of an area relative to the brightness of a similarly illuminated area that appears white or brightness relative to a reference emitted light that appears white. Chroma and saturation are both related to the amount of hue in a color or its colorfulness as compared to neutral (white or gray). Saturation by strict definition is colorfulness relative to the brightness of the 7 color, though at a constant brightness level saturation and chroma scales describe the same range of sensations (from pale or grayish to strong or vivid). Table 1.1 Perceptual and Related Psychophysical Color Terminology Perceptual Terms Psychophysical Terms hue dominant wavelength CIE hue-angle Munsell Hue brightness (luminosity) luminance lightness (value) luminance factor CIE lightness Munsell Value chroma CIE chroma Munsell Chroma saturation purity CIE saturation hue and saturation Chromaticity Adapted from Hunt (1987a) p. 71. The psychophysical terms prefaced with ’CIE’ (Commission Internationale de l’Eclairage) may be derived using either the CIELUV or CIELAB systems and the Munsell terms are for dimensions in the Munsell color order system (Hunt 1987a pp. 117-122). Cartographic Literature Many summaries of color use in cartography have described surround- induced changes in perceived color (Campbell 1991 p. 131, Dent 1990 pp. 384- 385, Imhof 1982 pp. 59-60, Keates 1982 p. 44, Robinson 1952 pp. 93-95, Robinson 1967 pp. 53-54, Robinson and others 1984 pp. 180, 182, Wood 1968 p. 56). These descriptions are used to warn that differences between map colors should be 8 large enough that map readers will be able to differentiate the colors regardless of variations in surrounding map colors. Limiting the required number of colors by limiting the number of map categories has been recommended to aid the designer in making robust color selections that are easily differentiated. Most authors also assured that outlines separating colors reduce induced changes. In addition to general descriptions of induction in the cartographic literature, induction on maps has concerned researchers of specific cartographic design problems. Literature on color charts, effective computer- displayed maps, continuous-tone choropleth maps, and equal-step color scales have provided discussions of the effect. Brown (1982 p. 111) and Castner (1980 p. 374) both recommended that cartographic color charts be designed with spaces between the colors to reduce induced enhancement of differences between adjacent colors on charts. Interviews on color chart use with cartographic designers (Brewer 1989-90 p. 6) supported the Brown and Castner recommendations. Four of the interviewees mentioned that difficulties with induced changes in color appearance occurred when using their charts, and two described their use of masks to isolate colors to reduce contrast effects. Castner (p. 377) also urged that color charts be designed to show only sets of colors that will remain distinctly different with changes in appearance caused by varied map surrounds. Spiker and others (1986) selected a set of point, line, and area colors for effective display of computer generated topographic maps and considered contrast effects a major variable affecting their recommendations. They used color naming to identify frequent confusions for symbol and background color combinations. They then used symbol search tasks to refine their color 9 selections and arrived at a set of color assignments that incorporated large contrasts to promote accurate identification. In contrast to the approach of Spiker and others (1986), in which identifiability is maximized, Ware (1988) recognized that form information (revealing the overall distribution pattern) may be as important as accurate color identification on thematic computer maps. His subjects matched continuous-tone map colors with legend colors, and he found that lightness and saturation scales were more susceptible to matching errors attributed to induced changes in appearance. Map-to-legend color matches with spectral progressions that show a full range of hues were less error prone. Ware recommended use of a color sequence that increases in lightness to reveal form and also includes hue differences to reduce induced confusions. Color differences are small and particularly susceptible to induced misinterpretation on continuous-tone choropleth maps. Peterson (1979) and Muller (1979) described the effect of contrast on their results in continuous- tone research. Muller presented subjects with the task of outlining regions of low, medium, and high population density on a continuous-tone map. He ranked counties using mean categorizations calculated from subject responses of 1, 2, or 3 for low, medium, or high (visual ranks). He also ranked the counties by their population densities (density ranks), which was equivalent to ranking by lightness. Muller related discrepancies between visual and density ranks to changes in appearance induced by the relative lightness of surrounding counties. Table 1.2 shows ranks and discrepancies for example counties for one of Muller’ 5 test groups (’non-geographers’). These mean displacements of counties within the range of densities on Muller's test map provide objective evidence of induction effects on map perception and evidence of induction as a confounding variable in cartographic research. 10 Table 1.2 Rank Discrepancies in Muller's Continuous-Tone Results Lightness of counties ranged from white at 1 (lowest rural population density) to black at 120. Equal rank differences did not imply perceptually equal steps between grays. Counties that appeared darker; surrounded by lighter, lower density counties: Visual Rank 73 92 95 100 110 Density Rank 69 87 91 94 107 Discrepancy +4 +5 +4 +6 +3 Counties that appeared lighter; surrounded by darker, higher density counties: Visual Rank 4 24 29 5 36 6 37 Density Rank 29 34 38 46 49 50 65 Discrepancy -25 -10 -9 -41 -13 -44 -28 (Muller 1979 p. 244) Cartographers working on equal-step scales have qualified their results with recognition of the problem of induction. Kimerling (1975 p. 126) states that background reflectance was the only factor significantly affecting the form of lightness-versus-reflectance curves in his research on gray scales. Cox (1980) attempted a systematic treatment of this problem, but he found a complex pattern of response with changes in background. He concluded that only perceptual differences between the lighter grays were affected by changes in background (p. 68). Robertson and O’ Callaghan (1986 p. 30-31) observed that the equal-step color sequences they developed need not be precisely uniform because induction substantially alters color perceptions. Kimerling (p. 126) suggested that equal-step gray scales can provide only general lightness-selection guidelines for the same reason. Similarly, Monmonier (1980 p. 36) and Castner (1980 p. 375) both cautioned cartographers to expect 11 imprecision in map reading partly because of induced changes in the appearances of map colors. A cartographic induction model would reveal the magnitude of imprecision that cartographers should accommodate in their design of map color combinations. Simultaneous Contrast Literature M.E. Chevreul is the ’father’ of simultaneous contrast research and made numerous controlled observations of the effect. Modern research results are consistent with his generalizations from 1839: (16.) All the phenomena I have observed seem to me to depend upon a very simple law In the case where the eye sees at the same time two contiguous colours, they will appear as dissimilar as possible, both in their optical composition [hue] and in the height of their tone [lightness and/ or purity]. We have then, at the same time, simultaneous contrast of colour properly so called, and contrast of tone. (17.) For two contiguous colours, 0 and p, will differ as much as possible from each other when the complementary of o is added to p, and the complement of p is added to o... (19.) when the colours are not of the same degree of intensity, that which is deep appears deeper, and that which is light, lighter; (Chevreul 1839/1981 pp. 50-51, italics are Chevreul’s) Chevreul specified the complementary color pairs as red-green, orange-blue, and yellow-violet, and he also considered black and white to function as a complementary pair. Birren noted that Chevreul’s complements are afterimage complements (Chevreul 1839/ 1981 p. 52). Brightness and Lightness Induction Since Chevreul’s time, a massive body of research on induced changes in brightness has grown and the topic continues to be reworked. Early research 12 is well surrunarized by Heinemann (1972), revealing a wide variety of functions describing results for different stimulus configurations and different luminance levels. Generally the original description remains acceptable: brightness induction causes a color to look darker with a bright surround and brighter with a dark surround. However, distinctions between brightness and lightness percepts are now better recognized and theoretical links between explanations of lightness constancy and brightness contrast have been suggested. These developments have lead to criticism of ambiguities in both test stimuli and instructions used in the study of brightness induction. Constancy, almost the obverse of contrast, is the ability of the visual system to maintain approximately unchanged perceptions with differences in both the level and hue of illumination (for example, the same piece of paper is perceived as white whether it is seen in bright daylight or the dimmer yellow of an incandescent lamp). A common component of constancy research conclusions is that equal ratios of surround and center luminance lead to equal perceptions of lightness, regardless of overall luminance levels. Contrast effects, on the other hand, are typically measured as discrepancies between absolute-luminance (brightness) matches rather than the relative- luminance (lightness) matches that may be judged by subjects. Therefore, the uncontrolled mix of lightness and brightness perceptions in induction- experiment designs contributes ratio-derived deviations in perceived brightness measures. Gilchrist (1988) designed an elegant demonstration of the effect of perceptual structuring on simultaneous contrast by using the same luminance arrangement within two contexts (Figure 1.1). A gray square (Munsell Value 2.1) was seen on the physically shadowed half of a white 13 ground. An array of grays, from which subjects were instructed to make a lightness match, was seen on the adjacent lighted half. In the constancy case, the shadow-casting structure was obvious to subjects and the darker surround was perceived as a change in illumination. Alternatively, subject views were masked and only simple dark and light backgrounds for the same array of gray samples were perceived. With the dark surround understood as a shadow, the median value match was 7.5 (much lighter than a 2.1 luminance match) and close to a correct match of paper samples irrespective of differences in illumination. Masking the view produced a median match of 3.0, which is a minor lightness increase compared to the constancy effect and is closer to matching the measured luminances under the different illumination conditions. Thus, with the same instructions, very different grays were matched with the shadowed test gray and the difference depended on interpretation of the display. Additional contrast and constancy research has focussed on the lightness- brightness distinction. Arend and Goldstein (1987) found large differences in selected matches with differences in instructions that specifically evoked evaluations of lightness or brightness with both simple and complex test stimulus arrangements. Evans, writing in 1948 (p. 166), had also found very different perceptions of simple center-surround stimuli with differences in instructions (these insights are therefore not confined to recent research). Iacobsen and Gilchrist (1988) replicate and adjust classic work on brightness contrast by Hess and Pretori from 1894. They found accurate brightness matches for stimuli brighter than their surround and fairly accurate lightness matches for stimuli darker than their surround. Thus, a gray with a surround of higher luminance tends to be judged relative to its surround and thus its lightness is evaluated more readily than brightness. The same gray 14 contrast condition constan condition . matches I matchescyn Test Data. 3 Values of Obsavers 1 I I I I u u u grays matched il l i l I i I I I l 1 1 r I to test gray 2.1 3 ‘ 5 6 7 a 3.5 within shadow equal luminance Mm”"v"“e equal ratio (brightness match) (lightness match) black paper casts shadow Constancy shadow Condition / comparison grays white paper Contrast Condition From Gilchrist 1988 p. 413 mask obscures view of shadow-casting paper Figure 1.1 Gilchrist’s Experiment Investigating Lightness Constancy and Contrast 15 seen on a dark surround will be perceived as luminous and its brightness is evaluated more readily than lightness. The experimental ambiguity of whether to judge lightness or brightness possibly explains Cox’ 5 (1980) interpretation difficulties with his gray-scale results from the cartographic literature. He may have found ”complex” differences and increased variation in the perception of only light grays with different background luminances because these light grays were compared by brightness, lightness, or both in an uncontrolled manner. The darker grays, however, may have been more consistently compared by lightness alone because they were frequently darker than their surrounds. The intent of the preceding review is not to imply that contrast effects do not occur (perfect brightness matching was uncommon in the controlled recent research reviewed in this chapter) but that test instructions and unconscious perceptual structuring of the display can radically affect results. It is unlikely that illumination would be interpreted as varying across a map sheet or across a complex map filling a computer screen. Therefore, the whites of the map may function as references for an assumed constant illumination and large shifts associated with local lightness constancy adjustments would not be perceived. In comparison, simple center-surround displays produce an ambiguous perceptual structure that may lead to inappropriate interpretations of test luminances. Therefore, research designed to solve problems of cartographic simultaneous contrast must test induction on maps, or complex map-like displays, and must include test instructions that evoke responses resulting from map-reading tasks. This insight renders much of the early research, based on ambiguous mixtures of lightness and brightness evaluations, of little relevance to prediction of brightness contrast effects in the cartographic context. l6 Hue Induction and Complementarity The research literature generally supports the traditional view that the appearance of a central hue is shifted toward the complement of the surround color (Jameson and Hurvich 1964, Kinney 1962, Krauskopf and others 1986, Troscianko 1977, Valberg 1974). With the exception of Jameson and Hurvich, these authors measured induced changes in Chromaticity (hue and saturation) of neutral-appearing centers, which provides little information on how perceptions of more saturated hues are affected by chromatic surrounds. A consistent deviation from complementarity was found by a number of researchers (Eichengreen 1976 tested a yellow center, Takahashi and Ejima 1983 tested a neutral center, Ware and Cowan 1982 tested 15 hues). Their results included a shift toward red with a blue surround rather than an increase in yellow, which is the expected complementary shift with a blue surround. These deviations suggest that induction effects between hues may be more complex than the simple complementary-shift summary implies. Another difficulty with the common generalization of complementary chromatic induction is that researchers work with many different systems of color specification and each system incorporates differing definitions of complementarity. Psychophysical complements are defined on the CTE-(x,y) Chromaticity diagram as colors that can be additively mixed, or averaged, to produce a neutral (Agoston 1987 pp. 68-70). Mixture lines are straight when plotted on the Chromaticity diagram and this convenient property places complements on opposite ends of any straight line through the neutral point (complements along straight mixture lines are also maintained on the CIE-(u',v') 17 chromaticity diagram used in Figure 1.2). Neutrality may be defined as the chromaticity of the equal-energy light source (E) for luminous-appearing colors with dark surrounds or at a standard daylight chromaticity for surface colors (CIE Illuminant C for example, Figure 1.2a). Takasaki (1969) worked with Munsell color samples, for which neutral has the chromaticity of Illuminant C, and his results show hue shifts parallel to Illuminant C complements. A viewer’s perception of neutrality is shifted when the visual system adjusts to the illumination of a scene (Agoston 1987 p. 188, Bartleson 1979). Adaptation is intertwined with simultaneous contrast but, in order to limit the scope of this work, I focus on studies that deal with changes in perception that are due primarily to simultaneous contrast, which is a phenomenon that occurs immediately on viewing stimuli. I have eliminated from discussion most of the research on changes in appearance with adaptation to various illurninants or stimuli (lengthy adaptation to a red surround for example). Perceptual complements on the CIE chromaticity diagrams (x,y or u',v'), however, are best defined using neutral as the chromaticity of the adaptation illuminant. For example, Kinney’ 5 subjects were adapted to incandescent light and her induced complementary shifts were well described by mixture lines drawn through the chromaticity point for Illuminant A (Kinney 1962 p. 512 and Figure 1.2b). Valberg’s (1974) work produces complements centered on a neutral point that is bluer than typical daylight chromaticities. Complement definitions may also be based on systems other than the CIE chromaticity diagram. Complements may be defined as the afterimage colors that appear after looking away from a color on which the eyes have been fixed (Agoston 1987 pp. 195-199 and Figure 1.2c). More loosely, complements are colors opposite on the color wheel, which depends on the chosen l l l l l 1.2a Additive Complements through 1.2b Additive Complements through Daylight Chromaticity (illuminant C) Chromaticity of Incandescent Lighting l l l ._ From Hill 1985 p. i4 &Aprton From Hut 1987'). I96 19670;). 123 l I l l l I 1.2c Afterimage Complements ‘l.2d Munsell Complements .6 \OW (“W \ s _ 50001-9 e _\¢r 700 ' ’7 Reds 650 D .. - ct“ 44 VI m ‘0 <9 9,” 3 - g 9‘ as 5 ‘0’ ‘ ‘3 2 8 From Hui was p. 14 J _ l l I ‘50 Fmel 1987a p.130, m 1.2e NCS Complemetary Unique Hues ° I I I I I 1 o .1 2 u'-3 .4 .s .6 1.2f Explanation of Diagram Structures Figures 1.2a to 1.2e show the boldly outlined portion of the above CIE 1976 (u',v') chromaticity diagram (1 .20. General hue-name regions are shown on this locator, and the curve plots chromaticities of the pure spectrum colors (wavelengths are in nanometers). The CIE (u',v') diagram is chosen as a common base for comparison of complements because it approximates perceptual scaling better than the CIE 1931 (x,y) chromaticity diagram but retains the property of straight-line additive mixtures. Figure 1.2 Comparison of Complementary Color Definitions 19 arrangement of hues. For example, the Goethe color circle sets the additive primaries red, green, and blue opposite their respective complements cyan, magenta, and yellow, which are the subtractive primaries (Agoston pp. 45-46). The Munsell color order system also provides a standardized hue circle used to define complements (Figure 1.2d). Unique red, green, blue, and yellow (also called psychological primaries or unitary hues) provide an additional set of complements. The hue circle of the Swedish Natural Color System (N CS) is one example of an opponent color system structured with unique red opposite unique green and unique yellow opposite unique blue (Hunt 1985 and Figure 1.2e). Hurvich (1981) provides a summary explanation of the opponent theory of color vision that he and Jameson have long worked to develop. A common characteristic of proposed opponent color mechanisms is that the pertinent dimensions used to explain hue perception are red-green and yellow-blue. Color opponency is not new to cartographers. Potential applications of opponent color theory have been examined in previous research by Eastman (1986) and Moellering and Kimerling (1990). I will discuss opponency further in the later section on quantitative models of induction. 20 Saturation or Chroma Induction The most direct way of modelling induced changes in colorfulness is to couple chroma with hue in an attribute of chromaticity that is subject to complementary induction effects. For example, a red of moderate chroma with a complementary green surround will appear to be a red of greater chroma, and the same moderate red with a strong red surround will appear to be lower in chroma, or grayed, because of mixture with the green complement induced by the surround. Thus, when the surround and affected colors are similar in hue, the induced complement will reduce the chroma of the center color, rather than cause a shift toward a different hue. Likewise, a color with a complementary surround will have a like hue induced and will therefore increase in chroma. Modelling chroma change by linking it only with hue induction assumes that changes in the brightness of the surround do not induce changes in chroma. Research relevant to brightness-induced changes in colorfulness has focused on saturation, which is defined relative to brightness, and the work has produced contradictory results. Classic rules of induced saturation, found in Kirschman’s ’laws’ from 1890 (Kinney 1962 p. 503), state that saturation of the induced color increases with increasing saturation of the surround and as brightness contrast with the surround decreases. Alternatively, Pitt and Winter (1974) and Kinney concluded that perceived saturation was greater with a contrasting bright surround than with a dark surround. Troscianko (1982) used color naming to measure induced changes (he defined saturation as ”the apparent amount of colour appearing to be emitted from the stimulus” p. 90), and he concluded that perceived saturation was greatest with 21 a neutral surround of near-equal brightness. He maintained that perceived saturation was decreased by induced whiteness or blackness from surrounds of contrasting brightness. Breneman (1977) and Valberg (1974) concluded that perceived saturation was independent of the brightness of the surround, though Breneman’s conclusions were for non-adjacent overall surrounds and Valberg worked only with a range of surrounds with brightness equal to or greater than that of his test colors. The range of conclusions about saturation changes induced by changes in brightness has ambiguous cartographic applications and further testing in map reading contexts would best clarify the relevant links between saturation and brightness. Test Methods Three general categories of method are used to quantify the magnitude of induced changes in perceived color: matching a comparison field, adjusting to match a unitary hue or white, and color naming. Matching tasks were most common in the research I reviewed. The typical induction experiment presents a central test color with a surround, or other inducing field, and a comparison color without the inducing field. Any one of the fields (test, comparison, or surround) may be adjusted to accomplish a perceived match between the test and comparison colors. Alternatively, a selection of comparison fields are offered from which to select a match. Use of a comparison field may be avoided by having subjects adjust the test color to maintain a white or to maintain a unitary hue (unique red, yellow, blue, or green). Color naming is a less popular method in which subjects are trained to identify, for example, the lightness, saturation, and hue of affected colors without comparison to another color sample. Response analyses in the three categories of method involve comparing differences between colors selected 22 as matches, comparing amounts of adjustment needed to produce a match, or comparing differences in descriptive names when colors are seen with different surrounds. Further variation in experimental method is found in modes of viewing test stimuli. Some researchers use haploscopic viewing in which one eye sees the test and the other eye sees the comparison color simultaneously. Haploscopic viewing is used in an attempt to prevent induction from affecting the perception of the comparison color, with the comparison color and surround-affected test color positioned adjacent to each other. Binocular viewing (in which both eyes see the same stimuli) is also used in induction research and is a method more applicable to map reading. The subject either looks back and forth between colors or colors are alternated while the subject selects a match. Spatial Aspects of Induction Spatial variables affect the magnitude of induced changes in color appearance. Induced change increases as the affected color decreases in size and as the inducing stimulus increases in size and decreases in separation distance. These generalizations are supported in almost all of the work in which spatial variables are tested. Conclusions vary on the spatial extent to which induction operates. In Heinemann’s (1972 p. 160) summary of early work, he reports that a separation of 1.5 to 2 degrees visual angle will prevent induced changes (1° visual angle is 1 cm at an arm’s-length viewing distance of 57 cm). More recent work of Blackwell and Buchsbaum (1988) produced a similar conclusion; minimal induced change occured with separations greater than 2° from a 0.6° test color. Leibowitz and others (1953) found that increasing the 23 separation of 0.5°-square test and inducing colors by distances up to 05° progressively reduced induced changes in perception. With their separations between 05° and 9°, induced perceptions were relatively unchanged. Whipple and others (1988) arrived at the same 0.5° separation cut-off for changing induction amounts with larger concentric-circle stimuli (centers ranged from 17° to 4.3°). Ejima and Takahashi (1983) found no further decrease in induction past approximately 0.3° separation from a small test color less than 0.1° wide. These results suggest that only the surround colors within between 05° to 2° separation have a substantial effect on perceptions. Note that the separation cut-off of 05° is much wider than an outline width, which cartographic texts suggest will eliminate or reduce simultaneous contrast effects that interfere with map color comparisons. Research work on surrounds contiguous with the central test color suggest there are induced contributions from more distant portions of the surround. Valberg (1974) found maximum induction effects on a 1°-by-2° test color with a 6° surround. Yund and Armington (1975) tested 40 size combinations of concentric centers and surrounds. They found continuously increasing induced change for increasing surround sizes (3° to 205° diameters) with constant center sizes (15° to 19°). They also found greater induced changes with decreases in center size. On the other hand, De Valois and others (1986) found no change in induction magnitudes with central stimuli ranging from 02° to 3° square with a constant surround size. They provide the only results contradicting the general expectations for size and induction relationships (their results may be peculiar to their stimulus that alternated between opponent complements). Again the results are contradictory. The spatial characteristics of induction should be considered in the prediction of map color appearances because map polygons can vary in 24 number, size, shape, and color in infinite combinations. Cartographic applications require a model of induction that is not dependent on a particular center—surround configuration. The most general induction model incorporating spatial variables is provided by Yund and Armington (1975). Their model predicts the color at a point using distances to edges (boundaries between different colors) for each opponent channel (red-green, yellow-blue, and white-black). Other authors propose less-developed models for complex surrounds that average distance-weighted colors in the surround (Ejima and Takahashi 1983, Marsden 1969). The general formula presented by Yund and Armington has the advantage of elegantly treating both center and surround sizes and shapes simultaneously. Their ”edge-distance” approach performed better than area and width measures in Yund and Armington’s analysis, though their tests were conducted with relatively large and simple concentric- circle stimuli. The emphasis of the Yund and Armington model on edges is consistent with psychophysical evidence that edges are important for perception and with physiological evidence that cortical cells respond to specific edge configurations (Livingstone and Hubel 1988). Color differences at edges are also critical elements of Land’s (1986) retinex theory that offers an explanation of color constancy. Edge processing issues in contrast and constancy explanations are well summarized by Gilchrist (1988). Quantitative Generalization about Induction Induction is often studied with the purpose of testing color vision theories. Early competition was primarily between retinal processing and opponent processing arguments. Application of results from this theory- motivated research in the map context is usually precluded because the theoretical formulations are too general, color measures are vague, formulae include parameters valid for only restricted ranges of stimuli, or test stimuli and viewing conditions are very different from map-reading contexts. Past tests of different models, however, do guide the way to successful approaches to induction prediction. Broad categories of induction models, which individual researchers have refined and combined, are summarized in the following paragraphs. Trichromatic theory relates color perception to the responses of three types of receptors: long-, middle-, and short-wavelength sensitive cones in the retina of the eye. Retinal treatments of induction are based on the von Kries coefficient law (from 1905) and explain induction using multiplicative constants that adjust the responses of the three receptors: C = f[(k1R), (kzc), (k38)] where C is the perceived color; R, G, and B are the long-, middle-, and short-wavelength sensitive cone responses to the test color; and k1, k2 , and k3 are constants that adjust for induction effects from the surround. Takasaki's (1969) research, for example, supported a von Kries approach to the prediction of induced changes in color. Opponent theory was put forth by Hering in the 1800’s and began as a competing alternative to trichromacy. Opponency mechanisms are now generally accepted as a second stage in the visual processing of color. 26 Opponent theory finds strong physiological support in the existence of red- green and yellow-blue opponent color cells 'in the lateral geniculate nucleus, the portion of the brain that initially processes short-, medium-, and long- wavelength responses from receptors in the retina (Livingstone and Hubel 1988). Opponent-process treatments of induction were pioneered by Jameson and Hurvich (1964) and have been supported in additional induction research (De Valois and others 1986, Ejima and Takahashi 1983, Kinney 1962, Shevell and Wesner 1989, Takahashi and Ejima 1983, Valberg 1974, Yund and Armington 1975). The basic opponent model hypothesizes that induced color changes are additive: C = f[(RGt + k1RGs), (“it + szBs), (WKt + k3WKs)] where C is the perceived color; RG and YB are the red-green and yellow-blue opponent responses; WK is the white-black luminance response; t and s subscripts designate the portion of the response attributed to the test color (t) and to the surround color (5) inducing the perceived change; and k1, kg, and R3 are constants that adjust the magnitude of induced change. A solely retinal explanation of induction has been effectively discounted in the research literature. The operation of induction effects over distances larger than retinal models predict and the effectiveness of dichoptic presentations (the surround is seen by one eye and the superimposed central test color is seen by the other eye) suggest that induced color interactions are produced at levels higher in the visual system than the retina (Whipple and others 1988). The lack of induced change in stimuli that alternated faster than approximately twice-a-second between black and white or between opponent colors also precludes a retinal-process explanation for induction effects (De Valois and others 1986). 27 Deviations from opponent-process model predictions have lead some researchers to suggest that adjustments at both the retinal and opponent stages of a two—stage color vision model better explain patterns of induced color change (Eichengreen 1976, Takahashi and Ejima 1983, Ware and Cowan 1982). Alternatively, Krauskopf and others (1986) rejected both retinal and opponent approaches and concluded that induction is a producet of interactions at still higher levels of processing in the visual cortex. Adopting an opponent approach, however, seems the most promising first step in the practical and fairly broad-brush approach necessitated by the varied conditions in which a cartographic model of induction would later be applied. Modelling Color Appearance The form and tractability of formulae used in attempting to model induced changes in appearance are sensitive to the units of measure used to specify color. The variety of available notation systems complicates the comparison of results and evaluation of function forms and parameters. The internationally accepted CIE-(x,y,Y) system was established in 1931 and is commonly used in the study of color induction, despite its lack of perceptual scaling. Some researchers have chosen to work with perceptually scaled transformations of CIE-(x,y,Y) that are named CIELUV and CIELAB and were standardized in 1976. Krauskopf and others (1986), for example, worked with color metrics defined by retinal receptor sensitivities. Others chose perceptual color order systems, such as Munsell, to provide precise color samples for use in their investigations. The importance of the chosen color metric has been recognized in induction research. In early work, Heinemann (1972) noted that failures of hypothesized additivity were not surprising given the arbitrary definitions of 28 induction magnitudes (p. 156). In work with induction and opponent theory, Jameson and Hurvich (1964) recommended use of units scaled by perceptual response (p. 151). For example, CIE I." lightness is a well-tested perceptually scaled measure that is a cube-root function of luminance (the latter is not perceptually scaled; a unit step at high luminance is perceived as a smaller change than at low luminance). Kinney (1962) expressed the difficulty of modelling induction before perceptually scaled color systems were becoming standardized: ”W hen functional relations break down, it is impossible to say whether this is due to a real phenomenon or an inadequacy of the system” (p. 518). Modelling color appearance is a more involved process than simply describing connections between the retinal and opponent mechanisms discussed up to this point. N ayatani and a group of other lighting and vision researchers have regularly documented the development of their color appearance model that focuses primarily on predicting appearances for varied states of adaptation (N ayatani and others’ 1990 paper includes a list of references for recent work on the model). Hunt’s extensive work on color reproduction has evolved into an appearance model as well (Hunt 1982, Hunt 1987b Chapter 8). Vision researchers are also working on models of the color vision system. The vision model under development by Guth and others (1980), for example, focused on threshold perceptions (color differences on maps are well above thresholds of perception). The purposes of models specifically focused on subjective color appearance, particularly the more general and applied Hunt model, are most likely to be successful frameworks within which to predict induction in the map context. The Hunt model is constructed in a physiologically plausible manner as a two-stage model. Chromaticity and luminance measures are first converted 29 to retinal receptor responses. Differences in these red-, green-, and blue- labelled magnitudes are then combined to produce opponent responses along red-green, yellow-blue, and luminance dimensions. The calculation output distinguishes lightness from brightness and saturation from chroma, and it accounts for adaptation given the chromaticity and luminance for a reference white (Hunt 1987b). The Hunt model produces perceptually-scaled appearance dimensions (lightness, redness-greenness, yellowness-blueness along with a number of other measures) that are good candidates for additional adjustment to account for induction from surround colors. The red-green and yellow-blue dimensions of the Hunt model become non-standard opponent dimensions once other attributes that affect appearance, such as adaptation, are accounted for. The adjusted opponent dimensions may be approximate but provide the following advantages: opponent complements aligned along continuous axes (unlike the NCS unique hue axes; compare Figures 1.3 and 1.2e), perceptually scaled color dimensions, and adaptation adjustments. Perceptual spacing is well recognized as a positive attribute for cartographic applications of color systems (Kimerling 1985, Robertson and O’ Callaghan 1986). Adaptation of the map reader to a particular illumination is unavoidable and needs to be accommodated if results are to be generalized to varied map reading conditions. Adaptation to the bluish white in a series of CRT maps or to the incandescent lighting on a paper map, for example, make expectation of induced complementary shifts with respect to only Illuminants E or C over- restrictive of a potential induction model. The Hunt model is still undergoing revision as other researchers begin to test it in varied applications. Therefore it has the disadvantage of not being a standardized system such as CIE-(x,y,Y) or CIELUV. On the other hand, 30 simultaneous contrast is a subjective effect that has not been well modelled in the objective realm of psychophysics and is better dealt with in the realm of subjective color appearance. Figure 1.3 Hunt Opponent Axes Hunt RG and YB axes plotted on CIE-(u',v') chromaticity diagram. 31 Summary Review of the simultaneous contrast literature reveals conflicting conclusions based on varied stimuli configurations that do not approximate the complexity of a map. Investigation of ambiguities in the treatment of lightness and brightness, as well as saturation and chroma, further support the importance of quantifying induction effects in a context approximating that in which the model will be applied. To establish a perceptual structure relevant to the map reading context, induction must be judged under realistic illumination conditions and test colors must be embedded in a matrix of other colors that includes reference whites. In addition, a test method that simulates color comparison during map reading requires binocular viewing and the rapid selection of appearance matches, rather than having subjects adjust a comparison color or be trained to name perceived colors. Review of past attempts to explain induced changes in color reveals that the opponent-process approach is the most common and promising avenue for modelling induction. The many perceptual influences on color perception beyond induced change suggest that previous research on other aspects of color perception should not be ignored in the pursuit of a successful model specific to simultaneous contrast. Thus, I have selected a color appearance model by Hunt that offers an appropriate and extensible quantitative framework for the induction prediction formulae that I proposed to develop. The model output is well suited to induction prediction in the map context: it is designed for application in a wide range of viewing conditions, is adjusted for adaptation, is perceptually scaled, and is opponent 32 based. Opponency is the basis of both the Hunt color appearance model and the Yund and Armington ”edge-distance" model that attempts to predict induction for spatially varied stimuli. Therefore, the two models may be linked by applying the opponent-based edge-distance formulae to the Hunt opponent color descriptions. Incorporating an opponent link between these two models within a final induction model is a promising way to accommodate spatial aspects of the inducing surround. Such accommodation is important to the prediction of induction effects with the varied spatial configurations of colors that appear on maps. CHAPTER 2 Experiment 1: Methods People were asked to match induction-affected colors to choices embedded in patchworks of gray hexagons for Experiment 1. This testing with complex but controlled displays produced data appropriate for modelling simultaneous contrast on computer-displayed maps. In this chapter, I describe and illustrate the test methods and sampling of color space used for the first experiment. In the next chapter, I will summarize and discuss the results of Experiment 1. In both Chapters 2 and 3, figures and tables follow the corresponding section of text in which they are referenced, rather than having figure and table pages interrupt paragraphs at arbitrary page breaks. Test Procedure Data collection for Experiment 1 was automated with SuperCard (Appleton 1990) programming running on an Apple Macintosh SE/ 30 computer with a 19-inch E-Machines color monitor. Subjects were tested individually with the test administrator seated left of and behind subjects throughout the session. Viewing was standardized using a chin rest that positioned subjects’ eyes approximately 57 cm from the display. Testing was conducted in a dim viewing area surrounded by a black curtain (ambient light reflected from a white calibration plate measured 1.5 cd/m2 with the monitor off). The black curtain and black cloth over the table and chin rest were used 33 34 to eliminate reflections on the monitor glass that could alter color perceptions in an uncontrolled fashion. A pilot test of 11 unpaid volunteers was conducted before final testing to practice instructing subjects and to watch for difficulties with the method. The pilot tests revealed that the sessions were too long but otherwise the test was running smoothly. The responses from these 11 subjects provided a preliminary look at data and allowed accurate editing to cut the experiment sessions to the final one-hour length. The pilot responses were not used in the final analysis. Appendix A provides the script for conducting an experimental session. Before testing began, an Ishihara test for color vision deficiencies was administered and the experimental task was explained using a series of ten practice displays. The practice sessions lasted between 3.2 and 10.6 minutes (mean 6.1), and during this time subject vision adapted to the general brightness of the computer display and illumination of the viewing area. Following practice, subjects responded to 100 different displays (out of 200 total) that were presented in a unique random order for each subject. Sixty subjects participated; thus 30 responses were collected for each display. A brief rest was incorporated in the testing session after every 25 responses. During this rest, subjects turned their attention to the test administrator and responded to background information questions (Appendix A). Subjects The responses of 30 men and 30 women with normal color vision were obtained in the experiment. An additional four subjects had been tested but were omitted from the sample (two had deficient color vision, the program malfunctioned for one subject, and one was omitted because of excessive 35 drowsiness). Subjects were recruited through posted advertising on campus and ads in the campus newspaper, and all subjects were paid ten dollars. I advertised for non-students and then accepted all who called (rather than encourage provision of false background information). Twenty-six subjects were students with the remainder from the wider Michigan State University community. Subjects ranged in age from 19 to 61 with a mean age of 26 years. Test Displays All test color combinations in Experiment 1 were embedded in a hexagonal grid (Figure 2.1). The center test color was surrounded by an inducing color, and there were six comparison colors in the outer display. These six comparison colors were arrayed in a random order for each display to avoid a positional bias in subject matching criteria. A patchwork of ten neutrals (listed in Appendix B) filled the remainder of the display. The first (white), second, fourth, sixth, eighth and tenth (black) of the neutrals were arranged immediately adjacent to the comparison colors to provide a surround that was varied and of medium lightness overall. The patchwork, out to approximately 4 cm around each comparison color, was constructed such that the surround of each comparison color was a rotated version of the same random pattern of grays, producing a constant comparison surround of medium overall lightness. One centimeter subtended one degree of visual angle on the test displays. The overall dimensions of the test displays were 25.1 by 25.7 cm. Three different densities of hexagonal grids were used with cells measuring 11, 6.5, and 4.5 mm between parallel edges (Figures 2.1, 2.2, and 2.3). The inducing surrounds ranged between 7.2 and 8.3 cm in diameter among the three grid 36 densities, and distances between the edges of the center and comparison colors ranged between 7.0 and 8.4 cm. With the test administrator pointing to the comparison colors on a practice display, subjects were asked to "choose one of these six choices that looks most similar in appearance to the center color." Center and comparison colors were marked by small arrows to make their positions clear. Subjects used a mouse to move the cursor to their choice and clicked the mouse button to record their selection. The measured color difference between the center color and the chosen match provided a measure of the induced change in perception of the center color caused by its surround. The click of the mouse button also cleared the screen and initiated presentation of the next display. The interval between displays was approximately 5 seconds as the next display was prepared, and during this interval a homogeneous medium- gray (30.5 cd/mz) display was presented to maintain subject adaptation. 37 Figure 2.1 Experiment 1 Test Display Format Hexagons of the largest size (11 mm) are shown at 50% reduction. Cell C is the test color subjects match to comparison colors positioned in cells numbered 1 to 6. The Surround, shown in black, induced changes in the perception of C. (N umber and letter designations were not present on the experiment displays.) Figure 2.2 Test Display of Medium-Sized Hexagons 6.5 mm hexagons reduced 50% 39 Figure 2.3 Test Display of Small-Sized Hexagons 4.5 mm hexagons reduced 50% 40 Most displays presented homogeneous inducing surrounds, but two deviations from this configuration were tested: surrounding all hexagons with outlines and breaking up the inducing central surround. As discussed in Chapter 1, cartographic texts have suggested that outlines will reduce induction effects on maps. Both black and white outlines were used with a subset of color combinations to investigate the differences in color perceptions that occurred (Figures 24 and 25). An additional subset of color combinations was tested with the inducing surround broken into random hexagons of a surround hue and mid-lightness gray, into black and gray hexagons, and into white and gray hexagons (Figure 2.6). These variations were tested to check whether or not induction was a significant influence on perception only in the rather artificial situation (with respect to maps) of a completely homogeneous inducing surround. 41 . '0- is . ‘i gag. 9% e»- k . g 155:5 3;. . :' .fi‘t‘fi es 9’. o Figure 2.4 Test Display with Black-Outlined Hexagons 42 mtr‘ (n. if x.) l C. ..r\ .._ww .1 x? Test Display with White-Outlined Hexagons Figure 2.5 43 Figure 2.6 Test Display with Example Heterogeneous Inducing Surround 44 Experiment 1 was designed to simulate conditions of computer-displayed thematic map reading. The use of a full-screen array of colors provided a complex choropleth-like context for the test colors. The comparison of colors within this complex graphic context was better representative of the comparison of colors within the body of a map than the simple center- surround stimuli often used in induction testing. The whites embedded in the patchwork surround provided reference whites that were important to the perception of lightness. The task of matching the center to a given set of comparison colors was similar to the map-reading task because a limited range of colors is used on a map and thus there is a limited set of colors among which the reader must distinguish. I felt that the alternatives of adjusting the comparison color to produce a match or of training subjects in color naming would sensitize subjects to the nuances of color difference to a degree that is not representative of the sensitivity brought to bear when reading a map. The importance of approximating the map-reading task in the design of the experiment outweighed the disadvantage of reduced precision with the limitation of preset selections of comparison colors. 45 Test Colors Color combinations were chosen to determine the shift both in distance and direction induced by a surround color. These shifts were measured in Hunt color space. An exhaustive sampling of color combinations from throughout color space would be prohibitively time consuming, and therefore combinations restricted to the opponent color axes were used. To quantify the magnitude or distance of color shift, reds were paired with greens in center-surround combinations, yellows and blues were paired, and a set of five grays were paired to sample the lightness dimension (Figures 2.7a and 2.7b). Off—axis combinations were included in testing the direction of change to ensure that the expected opponent directions of the shifts held with a wider range of hue combinations. A sample of eight colors, equidistant in color space from each center hue in a circular arrangement, were used for comparison colors for the direction trials (Figure 2.7c). Appendix B lists complete specifications for test colors: luminance and CIE 1931 chromaticity (L,x,y); Hunt red-green, yellow-blue, and lightness (RG,YB,J); and red-green- blue (R,G,B) for monitor display. The four center hues, at 1.0 unit along each opponent hue axis, were chosen to permit comparison colors of both higher and lower saturation and to permit distinction of the center from surrounds of like hue without an intervening outline or difference in lightness. The steps between the comparison choices were chosen to be equal in Hunt units and just noticeably different to my eye (I have normal color vision) once embedded in the patchwork of grays and at a distance from one another. For the distance trials, the six choices offered with each center-surround combination always 46 included a center match and the remaining five choices were colors to either side of the physical match with one or two more colors offered from the expected direction of shift. For the direction trials, six of the eight choices for each center-surround combination were offered with the two omissions from the direction of the surround with respect to the center, the least likely direction of shift. The direction trial comparisons were forced choice comparisons, since the physical match was not among the choices. Ten of the 200 displays were omitted from the analysis because of errors in color-table assignments that left crucial gaps in the sequence of comparison colors offered. The colors offered as comparisons for each center-surround combination are indicated with the data tables of Appendix C. The lightness of colors chosen to represent the red-green and yellow-blue axes was held constant (approximately 31.5 cd/m2 to produce a Hunt lightness of approximately 72.2). This level was chosen because it was midway between the lightness of the most saturated yellows and darkness of the most saturated blues that could be displayed on the color monitor. The mean luminance (31.3 cd/mz) of the gray patchwork surrounding the comparison colors was set to be approximately the same as the lightness of the test hues. Lightness was held constant for hue comparisons in order to facilitate generalization about induction along the opponent hue dimensions without the complication of lightness induction affecting perceptions simultaneously. Likewise, lightness testing was carried out with neutral stimuli to separate lightness induction magnitudes from hue effects. A potential disadvantage of the control exercised over the chromatic and lightness dimensions was that the medium-lightness hues featured in the primary testing were not of extreme saturation. The maximum saturation of these medium lightness hues was limited by the gamut of colors available at 47 that lightness on the monitor. Therefore, additional test displays were included that presented ranges of darker red, darker blue, lighter yellow, and lighter green that were more typical of saturated monitor colors (Figure 27d, Appendix B). These dark and light center hues were paired with surrounds of medium lightness that were the same lightness as the mean of grays surrounding the darker and lighter comparison colors. The medium- lightness surrounds were used in an attempt to make the lightness induction effects on the center and comparison colors equivalent while studying hue induction in these lighter and darker colors. Differences between comparison choices for the dark reds and dark blues were extended to 1.0 unit because differences of 0.5 were not discernable on the test displays (indicative of deficiencies in the perceptual structuring of the Hunt color appearance space). Hunt Axes 2.7a Lightness Test White light medium I Inducing surround colors dark II Center colors (also used as comparison colors and used as surrounds for opposite-hired centers) I Comparison colors (6 selected for each display) Black (K) 2.7b Hue Distance Test 2.7c Hue Direction Test . gl W light greens H_.+._._ G I I R _._,_._._._._. dark reds rd 2.7d light and Dark Saturated Hue Distance Test K .L._Y.'_._._. light yellows B I IV darkblues._._._. 4 : = bd Figure 2.7 Experiment 1 Sampling of Hunt Color Space 49 Production of Test Colors Production of monitor colors with specific Hunt coordinates (RG,YB,J) was a multi-step process. The luminance and chromaticity (L,x, ) of the aim colors were determined by repeatedly adjusting L,x,y used in the Hunt model calculations to produce the desired RG,YB,J (the Hunt model was not inverted). Once the L,x,y of aim colors were known, the R,G,B specifications of the monitor colors were adjusted to approximate the L,x,y aim as closely as possible. During this approximation process, colors were measured at five positions on the screen (C, 1, 2, 4, and 5 in Figure 2.1) with the gray patchwork filling the remainder of the screen as on the test displays. Colors were measured with a Minolta Chroma Meter CS—lOO, which is a hand-held colorimeter with through-the-lens viewing. I performed these color measurements positioned as a subject using the chin rest and seated in the test- subject location. The room was also set up with the same curtain and lighting as used for the test sessions to yield measurements of the colors as test subjects saw them. The Hunt model was run with a set of input constants derived for the particular viewing conditions of displays seen in the experiment. To specify color appearance, more information is needed than simply the L,x,y measurement of the color of interest. The input constant for the luminance of the reference white was 76 cd/mz, which was the highest luminance of whites measured on the monitor and was used so that calculated luminance factors (upper-case Y of the CIE 1931 system) did not exceed 100. The x,y chromaticity of the reference white (.270, .296) was set at the mean of measurements of all whites within the patchwork of neutrals for the largest- celled hexagonal grid. The luminance of the adapting background was set at 50 31.3, the mean luminance of the patchwork grays filling the screen. The conversion factor for calculation of the scotopic luminance of the adapting background was set at 1.2. This constant was derived using the color temperature of the x,y mean of the gray patchwork and extrapolating a value from example factors Hunt provided for 12 other color temperatures. The Hunt model was programmed in HyperTalk and explanatory documentation accompanies the code in Appendix D. Limitations on Color Precision Physical factors that may have affected monitor colors were location on the screen, other colors simultaneously displayed, fluctuations in power supply, aging of phosphors, amount of time the monitor had been on, and the previous color displayed at the same location. Commercial monitors are not precision display instruments and quantifying their vagaries was beyond my objectives. Steps were taken, however, to minimize these physical effects on the test colors. The monitor was turned on at least twenty minutes before testing began. The gray screen between displays was present long enough that the trace of the previous display had faded, and thus the same gray preceded each new set of test colors. Measurements of the colors on all test displays were made over the course of the period of testing to characterize the physical variability of the colors. Figures 2.8, 2.9, and 2.10 show graphs of measurements of the center test colors for all surround combinations and hexagonal grid sizes seen in the experiment. The variation in measurements is small compared to differences between comparison choices and is small given the generalized nature of my final guidelines for induction predictions. Thus, the coordinates of the aim colors were used in analysis of subject responses. 51 l ' ' jT | 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 l‘ -15 -14 13 12 11 -1 -0.9 08 ~07 ~06 -05 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 Y 111111 Figure 2.8 Center Measurement Variations for Medium-Lightness Hues Thin lines mark measurements of centers along segments of the opponent hue axes. Thick lines dropping below the axes mark aim centers at +/- 1.0. Adjacent comparison aim colors are marked by thick lines at +/- 0.5 and 1.5. 52 l t l 2.8 2.9 3 3.1 3.22 3.3 3.4 3.5 3.6 3.7 rd -1.8 -1.7 ~1.6 -1.5 -1 .35 -1.2 o1.1 .1 -0.9 gl 1.2 1.3 1.4 1.5 1.6 1.64 1.7 1.8 1.9 2 2.1 yl -4.7 -4.6 «1.5 -4.4 4.26 «1.2 -4.1 -4 3.9 -3.8 bd Figure 2.9 Center Measurement Variations for Light and Dark Saturated Hues Thin lines mark measurements of centers along segments of the opponent hue axes. Thick lines dropping below the axes mark aim centers. Adjacent comparison aim colors on the green and yellow axes are marked by thick lines 0.5 units from the aim centers. For the two dark hues (rd, bd), comparison colors are 1.0 unit from centers, beyond the axis segments shown. The sequences of colors at light and dark levels were chosen by finding the lightest and darkest saturated opponent color that the monitor could produce. The even-stepped differences between choices were then calculated in toward neutral from these maximum saturation positions. Thus, the positions of the test hues do not fall at ’nice’ number positions such as 0.5, 1.0, and 1.5 as with the medium lightness test hues. Thin lines plot measurements of the lightness of centers only (light, medium, dark) along the J axrs. Thick bars mark center- and comparison- lightness aim positions along the lightness scale. 53 80.3 74.5 — 62.9 ll Ill ”I ll l 57.1 lllll [HI] _ 51.3 -b— 45.5 lJlJl I II" 39.7 I J l l l — 33.9 light medium dark Figure 2.10 Measurement Variations for Center Lightness 54 The impact of physical color variation on subject matches was characterized along with subject variability. All colors were judged within control displays that presented the same patchwork of grays around the center color as the patchwork around the comparison colors (Figure 2.1 1). These controls for three sizes of hexagons produced data on the accuracy of matches that subjects made given the inherent variability of monitor colors without the effect of induction from the central surround. Color variations that were small relative to differences between comparison colors were not considered of importance to the final conclusions. Expectation of tight control of computer displayed colors for application of the model developed would render the work of limited use as a design aid to other cartographers also working on commercial monitors. Thus, I preferred incorporation of this physical variability into my analysis, just as variation within the group of 60 diverse subjects provided an approximation of the variability of map readers’ perceptions. 55 Figure 2.11 Experiment 1 Control Display Format CHAPTERS Experiment 1: Results and Discussion Recall that the overall objective of my research was development of a quantitative model of induction to aid selection of sets of easily identified map colors. Rather than predict average color perceptions for particular center-surround color combinations, I acknowledged the inherent variability in map readers’ perceptions of color by developing generalized perception buffers that accounted for at least 90 percent of test subject responses. The task of selecting colors that will not be confused once they appear with numerous surrounds on a map thus becomes a task of selecting colors that do not have buffers that overlap in color space. Application of the model requires calculation of induced-change buffers for all combinations of colors on a map. These buffers describe regions that encompass most map readers’ potential perceptions. In this third chapter, I will describe the results of Experiment 1 and discuss the analyses used to develop the induced-change buffers. Experiment 1 data were compiled as frequencies of choices for each comparison color matched with the center colors (Appendix C contains a complete listing of the response-frequency data). Distributions of frequencies were compared using both chi-square and Somers’s dyx non-parametric statistics. In addition, ordinary least-squares regression of center-surround contrasts against mean induced perceptions was used in model construction. The hypotheses driving the use of these statistics will be detailed as results are described in the sections of this chapter. Because multiple questions are 56 57 addressed with this data set, I chose to place discussion with the results summaries within each section, rather than in a section remote from the relevant tables and figures. The Somers’s dyx statistic measures association for two ordered discrete variables. The calculated value of the statistic ranges between positive and negative one, and these extremes indicate perfect prediction of the dependent variable by the independent variable. Zero indicates no relationship (i.e., no dependence of the row variable on the column variable). In the Experiment 1 data, subject choices (rows in all tables) are hypothesized to be dependent on differences in surrounds (independent column variables). The asymmetry of dyx is suited to the questions I ask of the data: to what extent does knowledge of the surround help predict subject matches? Asymmetry in a test means that the independence and dependence affects the calculation procedures. Asymmetry distinguishes dyx from other nonparametric measures of association (such as gamma and tau) that are symmetric. The asymptotic standard error for the statistic is used to evaluate whether the calculated dyx departs significantly from zero. Somers’s dyx is a PRE (proportionate reduction of error) type of measure requiring systematic comparison of every pair of responses. The statistic is described in Somers 1962 and 1980, Agresti 1981, and Bohrnstedt and Knoke 1982. The chi-square test of independence was also used to evaluate whether frequency-response distributions for pairs of color-combinations were significantly different. Chi square was included in the analysis to provide confirmation of results from the less familiar dyx. Chi square, however, is not sensitive to the ordering of categories and has the disadvantage of minimum expected-frequency requirements necessitating the collapse of row categories. Size With the exception of De Valois and others (1986), previous researchers found that induction produces greater changes in appearance for smaller color areas (Chapter 1). To investigate variation in the strength of induction with size, I tested center hexagons of three sizes: 11, 6.5 and 4.5 mm in diameter (1.1, 0.65, and 0.45 degrees visual angle). The most pronounced differences in induced change were predicted to be between the 11 and 4.5 mm hexagons. Table 3.1 shows the Somers’s dyx for 19 hue and lightness distance tests, indicating the extent to which size allows prediction of response frequencies. Appendix C lists the response frequencies for individual hexagon sizes. In size comparisons for 15 of 19 color combinations, the smallest hexagon size was not associated with significantly greater frequencies of shift than the larger hexagon size as had been predicted (Table 3.1). Chi-square tests produced the same pattern of significance for differences in frequency distributions. Though not significant, a greater shift in appearance was associated with the larger size for the y B and b B combinations and for grays with dark and black surrounds (these are shifts opposite of expected results). For four red-green combinations, the smaller hexagon colors underwent significantly greater change in appearance than larger hexagons with the same surround. Given the lack of significant size effects for the remaining 15 comparisons, however, data for the three sizes were aggregated for the remaining analyses. Conclusions will be applied to the overall range of sizes represented in the testing. Thus, the promising ”edge-distance” model of Yund and Armington (1975) that was described in the Chapter 1 review will not be used. 59 Table 3.1 Somers’s (1,, Comparison of Change Induced in Small and Large Test Centers Each N=60 Hue Distance Test Colors: Centers r g y b Surrounds adjacent R -.06 G -.38 “ Y -.12 B +.20 opposite g +.59 “ r +27" b +.13 y +.15 G +.33" R +.18 B -.05 Y +.21 Null Hypothesis: small hexagon sizes are not associated with a significantly greater induced shift in perception “ Reject H0 at .01 confidence level, one-tailed " Reject H0 at .05 confidence level, one-tailed positive (1 x: small size associated with greater saturation than large size negative cl'yx: small size associated with greater neutrality than large size Lightness Distance Test Colors: Centers 1 m d Surrounds W -.03 W -.03 d -.14 l -.20 K -.07 K -.18 K -.11 Null Hypothesis: small hexagon sizes are not associated with a significantly greater induced shift in perception Fail to reject H0 at .05 confidence level (one-tailed) for all lightness combinations negative dyx: small size associated with greater darkness than large size Controls Color matches with the gray-patchwork control surround offered an important baseline for the analysis of Experiment 1 results. The control results were used to determine whether subjects could accurately match centers with comparison choices in the absence of an inducing surround. The frequency-response distributions for the controls also offered an objective distribution with which to compare the perceptions of the colors with inducing surrounds. This comparison is made to answer the question of whether or not shifts in perception were significant. Figure 3.1 provides a summary view of the control responses. Table 3.2 lists the dyx for responses to colors seen with an inducing surround compared to the corresponding control responses (responses for three hexagon sizes were aggregated). The accuracy of the control matches was surprisingly good given the separation between the center and comparison colors of approximately 7.7 cm and the small color differences between choices. The minor physical variation evident in measurements of the colors (Figures 2.8 and 2.9 in previous methods chapter) did not hinder subjects’ abilities to match colors accurately to the comparison choices. Steps between lightness choices, however, were smaller perceptually than differences between hue choices (their numerical units are differently scaled), and lightness differences are generally more difficult to distinguish than hue differences over a distance (Hunt l987a p. 118). These difficulties manifest as reduced accuracy in matching the control grays; 79 percent of matches were accurate for the lightness control displays (N =270), whereas 95 percent of hue control comparisons were accurate (N =360). Establishing a buffer of one step (5.8 J 61 units along the lightness axis) to either side of the center lightness accounted for 98 percent of the control matches. This matching tolerance was used to evaluate induced shifts in lightness perception that are discussed in the next section. Most of the 32 color combinations listed in Table 3.2 were associated with significantly more choices of comparison matches from the predicted directions of perceived change than for the corresponding control cases. The exceptions are seven of the lightness and saturated-hue combinations. The centers for three of the four lightness combinations with light (1) and medium (m) surrounds were not perceived as different than the same gray center with the control surround. The differences in lightness between 1, m, and the control surround are interpreted to be small enough that no significant differences in perceptions were induced (lightnesses are 74.5 for the l surround, 57.1 for the m surround, and a mean of 61 for the control surround of six grays immediately adjacent to centers). The four light and dark saturated-hue centers with opposite surrounds show induced shifts in the reverse of the predicted direction. I discuss these deviations in the later section on the saturated hues. Chi-square analyses yield the same pattern of significance for the controls with the exception that the four saturated-hue ’opposite’ combinations had significantly different distributions of frequencies. The frequency data for the controls are listed with the relevant frequencies for induction sets in Tables 3.3, 3.4, and 3.5 that appear in later sections. Figure 2.7 shows labels for color-space axes. 62 W W 211 76 light 4 2 median ll 9 1' rs dark 13 K K 7. matched to gl center 8' 4923 Percentage of responses for light, medium, and dark centers with control surround Percentage of responses for medium-lightness red, green, yellow, and blue centers with control surround Percentage of responses for dark red, light green, light yellow, and dark blue centers with control surround Percentage of - g - ' Shawl:- 1: a center 7. mdched to 11 center 11 8 3 yl _ '5' _ 2 14 7; 10 % mlched to H center Figure 3.1 Comparison-Color Choices Plotted in Hunt Color Space for Control Surround Test 63 Table 3.2 dyx Comparison of Change Induced by Control and Inducing Surrounds Each N=180 or 150 Hue Distance Test Colors Centers r Surrounds adjacent R -.93 “ opposite g +.48 “ G +.56 “ G -.97 “ +.33 “ R +.19 “ H Lightness Distance Test Colors Centers 1 Surrounds -.74 " +.16 * W l m +.09 d K +.6l " m -.24 “ +04 +71 “ +.75 “ -.94 “ +.26 " +.29 “ ”6‘4 (1 -.27 “ -.19 " +.13 +.60 ’“ Light and Dark Saturated Hue Distance Test Colors Centers rd Surrounds adjacent R -.80 “ opposite G -.08 31 G -.77 u R -.26 yl Y -.83 “ B -.13 «‘5 B Y -.99 “ +.39 “ +.33 “ bd -.51 “ -.25 Null Hypothesis: In comparison to control surround matches, contrasting surround matches are not associated with a significantly greater induced shift in perception in the predicted direction. “ Reject H0 at .01 confidence level, one-tailed " Reject H0 at .05 confidence level, one-tailed positive dyx: inducing surround associated with greater saturation or greater lightness than control negative dyx: inducing surround associated with greater neutrality or greater darkness than control Lightness Test Induced lightness change is more difficult to measure than induced hue change because the comparison grays that are offered must also have surrounds with some lightness component (with hue induction, a comparison hue can have a neutral surround with no hue component). A black surround cannot be considered as ’no’ surround because it does induce a change in perception. For analysis of the research results, both the center grays and comparison grays were considered to have inducing surrounds (the comparison surround was a constant of 61 J, the mean of the patchwork of grays immediately adjacent to the comparisons). Table 3.3 provides lightness response data. Yund and Armington (1975) found center-surround contrast to be a strong predictor of surround induction, and opponent-theory approaches to induction generally postulate an additive effect (i.e., the induced effect is added to the perception). Therefore, to model the induced lightness perceptions, a proportion (k) of the lightness contrast between the gray test center (t) and its inducing surround (s) was added to the center gray (t). The same proportion (k) of the contrast between the lightness of the perceived match (p) and the mean of the gray patchwork (c) around the comparisons was added to the perceived match (p). This contrast relationship was consistent with previous research and produced the following equation: t+k(t-s)=p+k(p-c) (1) The proportionality factor (k) was estimated as 0.135 with ordinary least- squares regression. Rearrangement of Equation 1 clarifies the appropriateness of estimating k using regression with a forced intercept of zero: t-p=kl(p-c)-(t-s)] (2) The independent variable on the right of Equation 2 represents the difference in center-surround contrasts for the test color (t) and its perceived match (p). The variable dependent on this contrast difference is the induced change in perception, on the left of Equation 2. The mean of all matches to the test center was used as a surrogate for the perception of the induction-affected test color (p). Figure 3.2 presents a scatterplot of the data with dependent and independent variable axes. The line on the scatterplot shows predicted perceptions. It has a slope of 0.135, the k estimate, and passes through the origin. The r2 for this relationship is .86, a satisfactorily high percentage of explained variance. The slope coefficient is significant at the .001 confidence level (an intercept that is not significant is produced if the line is not forced through the origin). Once k was estimated, Equation 1 was solved for p, the predicted perception: p=[t+k(t-s+c)]/(l+k) (3) (t is test center lightness, s is inducing surround lightness, and c is comparison surround lightness). Figure 3.3 shows 12-unit buffers (needed with the control comparisons) centered on the predicted perceptions of lightness. The overall number of responses that these shifted buffers encapsulate accounted for 90 percent of subject lightness matches (N =870, control matches were not included). The only substantial outliers in this process were 14 subjects (22 percent, see Table 3.3) who perceived the light center with a white surround as darker than predicted. Each predicted match corresponds to at least 13 percent of responses for the individual color combination (those response frequencies that fall within buffers). This rate indicates a lean model that does not rely on extravagantly large buffers for its high percentage of successful prediction. 66 Table 3.3 Percentage of Matches to Comparison Choices for Lightness Distance Test Matches to I , light center Surrounds: W ( m d K (lmiu‘s‘: +1112 (control) U 411.1. lighter + 1.11 2 20 32 37 58 25 76 62 55 18 lll.ll( h -S.ll (Lulu-r (control) 13 370 63 Surrounds: (1101((‘SZ +232 +174 ‘ lighter +1111 792041 4660767029 39 2813 8 3 67 Induced change in perception (t-p) lfifi r l ' l ‘ l ' I ~60 ~40 ~20 0 20 40 60 Difference in center-surround contrasts (p~c)-(t-s) Figure 3.2 Scatterplot for Mean-Response Lightness Data Line has slope 0.135 and intercept at the origin. Surrounds: White light control medium dark Black 7 W 18 20 32 37 58 *1 25 -l l 7+4 62"-l ssl’d 13 -llght 50l- 4 13 ‘ 3‘ 70 I 63% control 56 —m ”ll-m 87,--m m ZSr-m 20 medium 32 " 13 s 23 9 20 41 } 46 -d GOP-d 76ll-d 70+1—d kd 29 -dark 39 " 28 13 ll x Figure 3.3 Lightness Induction Prediction Buffers Each of the six bold vertical lines represent the lightness axis. Tick marks to the right of each axis mark center lightness positions and black squares mark surround positions. The numbers along the left of each axis are percent responses for comparison choices (compare to Table 3.3). Bold ticks to the left of axes show predicted mean perceptions, and thin vertical bars represent buffers needed to account for 90 percent of perceptions overall. The percentage of perceptions an individual buffer encapsulates is the sum of the adjacent percentage figures. For example, at the lower left, dark gray with a white surround is perceived as darker and the 12-unit buffer includes choices selected by 46 and 39 percent of subjects. 69 Hue Distance Test Patterns in the medium-lightness hue response data mandated rethinking plans to treat the red-green and yellow-blue axes as continua with functions similar to that developed for lightness. Estimation of k from Equation 2 with center-surround hue contrasts and mean perceptions offered explanation of only 61 percent of variation in the data. Figure 3.4a shows a scatterplot for hue of the same form as Figure 3.3 for lightness. Examination of departures from the best-fit line revealed systematic errors (Figure 3.4b shows each point named). Note that perceived change for some of the extreme values of the independent contrast variable collapse back toward zero induced change. These extremes were mean perceptions of centers with surrounds from the opposite ends of the opponent-hue axes. The magnitudes of induced shift were closely linked to whether the surround hue was opposite or adjacent the center hue in color space. Examples clarify the adjacent and opposite terminology I use: yellow surrounded by blue (y B or y b) is described as an opposite hue combination, and blue surrounded by a more saturated blue (b B) is described as an adjacent combination. On the scatterplot, the eight points hovering between 0.1 and 0.3 above and below the zero-change line all plot perceived changes for opposite center-surround combinations. The colors undergoing mean shifts greater than 0.6 each had adjacent surrounds of like hue. These patterns of perceived change are seen in Figure 3.5, in which percent responses for each comparison color are plotted in color space (Table 3.4 lists these percentages). Comparison of frequency distributions using dyx revealed that the number of shifts in perception induced by the more- and less-saturated opposite 70 surrounds were not significantly different (.10 dyx for r g compared to r G responses; .02 forgrtogR; -.13 foryb toyB;-.06 forb ytob Y). My sampling of color space was not sufficient to model each neutral-to- hue range as a continuum, so additive constant shifts for opposite and adjacent center-surround cases were derived. A one-step buffer that extends to one step (0.5 units) of greater saturation from the color accounted for 95 percent of subject matches with opposite surrounds (N =690). A one-step buffer to neutral that extends from one step toward neutral accommodated 97 percent of subject matches for adjacent surrounds (N =360). Note that the greater magnitude of the shifts induced by the adjacent surround compared to effects of either opposite surround is evident in the dyx values in Table 3.2. Adjacent combinations produce stronger dyx than opposites (more than l0.9| versus less than l0.6 I) when responses are compared to frequencies on the control tests. 71 3.4a 1 E I c 0.5 -' .2 . O- «1 § . 3_ .. .s o - 3° I 5 s I U 4 '§ .05 4 3 u E . I . ' -1 l I I -4 ~2 0 2 4 Difference in center-surround contrasts (p-0)-(l-s) 3.4b 1 q ‘ ad'acent ’5. ' center-wr’round y-Y 3', " combinations 'r R r: 0.5 q .2 .. Q -1 . posite g centerosurround :2! b Y- 8' l combinations g R o 5 0 ‘ =controls % ‘ I y B y'b I. opposites 2 '4 n r C 5 U .. B -005 - a 1 '2 1 c - ‘ b s '- -I 0 adjacent: -1 l I l -4 ~2 0 2 4 Difference in center~surround contrasts (p—0)-(t-s) Figure 3.4 Scatterplots of Mean-Response Hue Data 72 Table 3.4 Matches to r, the red center Surrounds: adj.“ enl opposite (onlrol R g ( hoit es: more red malt II grater .. . neutral -1 Surrounds: adjacent opposite control C r 111.111 h g K'I‘Y‘ green +5 neutral +1 Red- and green-surround induced shifts in color space xmached [Y logcenter 23 73 :1 {-r R = Y = I e-g as so 7 bl Vsmatched lorcenter Vamlched Y togcenter Percen eof s 32 r2 1 r-é “'3‘: G 39 1 42 49 s llMode $111:ch limbmode b torcenter Heisman“ Figure 3.5 Hue Distance Test Percent of Matches to Comparison Choices for Hue Distance Test Matches to y, the yellow center Surrounds: adj.“ ent opposite control (7111111 1": "luff yellow 111.111 11 grayer -.' neutral - Matches to b, the blue center Surrounds: adjacent opposite ( onirol B y Choices: more blue malt h h 'm er +5 1. ) neutral +1 Yellow- and blue-surround induced shifts in color space Y 1 9L matched to y center '4 matched 1 1 Y ‘4 Y 69 .1, 2 +30 —0— 8 3 r 3 1 1s ' b 33 b 1 6 so 1. matched 3 to 5 center 7. mached to 5 center | s Comparison-Color Choices Plotted in Hunt Color Space for Hue Direction Test The direction test of hue shifts offered comparisons 0.5 units in hue difference in six directions in color space from each center color. Expected perceptions, based on opponent theory, were toward the opponent complement of the surround, and Table 3.5 is organized such that the number of matches to the expected perception are arranged across the middle row of each table. Matches counter-clockwise and clockwise to either side of the expected response are arranged above and below this central row. More extreme responses were aggregated and are listed in the top and bottom rows of the tables (Appendix C contains full data listings). The direction data are plotted in color space in Figure 3.6. Looking across this set of diagrams shows that shifts were generally away from the surround color positions. The analysis of lightness and medium-hue distance data made successful use of modest buffers that encapsulated high percentages of subject perceptions. In keeping with this approach, a fan-shaped buffer spanning 90 degrees in the opponent-complement direction from the surround was used to account for 90 percent of subject matches (N =930). Figure 3.6 shows the perception buffers for all center-surround combinations. Note that the modal-response frequencies are all contained within these buffers; the application of predicted opponent-complement shifts is well fit to these response data. This success confirms that offering comparison colors along only the opponent axes was appropriate for the hue distance test discussed in the previous section. The difference in magnitude of shift for opposite and adjacent surrounds for the distance test is also revealed in aggregate direction data of frequencies 74 for opposite and adjacent surrounds (N =120 for opposites and N=120 for adjacents). For the adjacent surrounds, 83 percent of the matches were the expected opponent complement with no responses straying outside the 90- degree buffer. In contrast, only 36 percent of responses were matches in the expected directions for the opposite surrounds with 27 percent of responses scattered outside the buffer. Greater variation in responses occurred for opposite surrounds with no clear alternative to the opponent complement for the direction of shift. Looking back at the distance-test response frequencies for these opposite surrounds (Table 3.4), 60 percent of subjects perceived an accurate match, unaffected by surround induction. The direction tests were forced choices (no accurate appearance match was offered if no induced change was perceived). Thus, greater variability of responses was consistent with randomness provoked in subject responses when no acceptable match was available. Greater consensus was apparent for adjacent-surround matches in the direction test and, likewise, 97 percent of subjects selected matches at least 0.5 units toward neutral in the corresponding distance tests. 75 Frequencies of Matches to Comparison Choices for Hue Direction Test Table 3.5 Each N=30 (note that tables list frequencies, not percentages) Matches 1.0 ' Surrounds: 31de cut ”Pl’mi": red center ( hoices: ( ounler- 1 Iotlmise >+43 from expet led +41 expec led ('01 hWiSt‘ ~43 from expected (4') MatChes to ‘ Surrounds: Adj-It enl green €871le BC Choices: counter- clockwise >+4S from expected +45 expected clockwise -4S ‘ from expected <45" Malc'l'es w Y Surrounds: adiarffll °PP°‘“*’ yellow center Matches to b blue center (‘ltoic es: counter- clockwise >+45 from expected +45 expected clockwise -4’i from expected <-45 Surrounds: (’ltoices: counter- 1 lot lux ise >+43 from expected expected c loc luvise -43 from expected ("1% YR Y CV BC adj.“ on! opposite YR 76 5:55., 4:“ a} I I o g 0 0.2.4- ............... . Surround color a Frequencymodeolcolor matches to center I Centercolor I Colormatchedbymorethantwosubjects ggg Paceptionbuller . Colormatchedbyoneortwombjects Figure 3.6 Comparison-Color Choices Plotted in Hunt Color Space for Hue Direction Test Light and Dark Hue Distance Test The light and dark saturated-hue test produced puzzling results. As with the medium-lightness hues, the shifts in perception were stronger for adjacent than for opposite hue combinations (note values of dyx in Table 3.2 and percent responses in Table 3.6). The opposite-surround shifts were all toward reduced colorfulness (the direction opposite to that predicted) though only two (g1 R, bd Y) were significant in comparison with control frequencies (if the alternate tail for one-tailed significance was used to evaluate dyx). Adjacent-surround shifts occurred in the expected direction of reduced colorfulness. The saturation of the dark hues (bd and rd centers) was slightly greater than the saturation of their medium-lightness adjacent surrounds (Figure 3.7). This relationship points out a contradictory case for the two commonly made generalizations about induction: the complement of the surround is induced and greater difference between the center and surround is induced. Accord with the second generalization requires these dark hues to shift in appearance away from their surrounds, not toward them with addition of their surround’s opponent complement. These bd and rd centers also point out an omission in my sampling of center-surround combinations. I did not test the effect of surrounds that were the same hue as the center and less saturated than the center. An alternative interpretation of the responses for these two dark hues does not produce a contradiction of the generalization that induced shifts increase the perceptual difference between the center and surround hues. The hue axes may be disproportionately expanded at low lightness in the Hunt color space. The suspicion that the axes were perceptually stretched for 78 darker hues followed from the decision to select comparison hues 1 unit in color difference to allow discernability (rather than maintaining the 0.5 unit difference used for selection of medium and light comparison hues). This interpretation suggests the interesting possibility of using induction to aid the difficult evaluation of relative saturation at different lightness levels. Lightness differences between center and surround colors reduced induction magnitudes overall. At medium-lightness, a similar effect may have been the greater strength of induction by adjacent (like-hued) surrounds compared to induction by opposite-hued surrounds. The lessened induced change in saturation with additional qualitative differences (differences in hue, differences in lightness) suggested that distinctions by saturation alone may be particularly susceptible to induced shifts in appearance. Because of the lessened induction effect with lightness differences, the overall prediction guidelines in the Chapter 3 summary recommend that perception buffers for adjacent hue combinations extend 1 unit from the original center color. The hue-distance analysis supported a narrower prediction that did not include the center color. 79 Table 3.6 Percent of Matches to Comparison Choices for Saturated Hue Distance Test Matches to rd, dark red center V. matched Surrounds: adj. opp. to gl center control It (i 1 72 27 I s I It _ rd H 'O' 79 19 x mached to rd center 9‘ matched to gt center 1 19 79 1 Surrounds: l G I 8 control '5! - H 11 M Q 9‘ matched to rd center Percentage of Responses: fl Mode I 10% to mode ' lees than 10% is matched ‘ to yl center Surrounds: .Irlj. opp. 22 70 8 control ' B yl I Y bd G 9 58 Y. matched '5 matched b H center to yl center 1 29 60 10 yl .Idi. opp. c onlrol Y 6 i mached to ltd center Figure 3.7 Comparison-Color Choices Plotted in Hunt Color Space for Saturated Hue Distance Test 80 Variations on Displays A subset of displays in Experiment 1 were varied to test secondary hypotheses about center-surround interactions at single grid densities. As described in the previous methods chapter, both black and white outlines were drawn around all cells in the hexagonal grids used to build the test displays (Figures 2.4 and 2.5). The heterogeneous display variation described in Chapter 2 involved imposition of a pattern of gray hexagons within the inducing-surround colors (Figure 2.6). For example, the heterogeneous version of r R presented saturated red and neutral (N) of the same lightness around the lower-saturation red center. To examine the effects on perception of these varied displays, response data were compared with responses to the standard versions of the displays (contiguous and homogeneous inducing surrounds with the same hexagonal grid densities). Responses were also compared to corresponding control responses. Each comparison involved 60 responses. The data for three outline comparisons and six heterogeneous combinations were omitted from the analysis. For these combinations, the distinction of inducing versus control surrounds did not produce a significant difference in response prediction for the single hexagon density (recall that responses for color combinations seen with two or three hexagonal grid densities were aggregated for the primary control comparisons). The omissions were made because this discussion of results focuses on whether responses to the display variant were more like the control or more like the induced responses. If there was only a weak difference between the control 81 and induced response patterns, then additional variants of that display combination had nothing to contribute to the discussion. Hexagon Outlines Table 3.7 presents dyx statistics for comparisons of outlined-hexagon responses to the contiguous inducing-surround responses and to the control- surround responses. Among 1 7 comparison pairs, there are only three for which outlines caused the responses to be not significantly different from the control responses and significantly different from responses to the standard inducing display (black outlines for r g and white outlines for rd R and bd Y). For an additional two comparisons (white outlines for g G and m K), outline versions produced responses that were significantly different from both the control and inducing cases. The intermediate nature of the responses for gl R (white outlines) rendered them not significantly different from either the inducing or control responses. These last three comparisons revealed an intermediate reduction of induction effects by the hexagon outlines. The remaining eleven dyx comparisons show that knowledge of whether hexagons were outlined or contiguous provided no significant improvement in prediction of the subject responses. For the majority of comparisons, the addition of outlines did not reduce the amount of change induced by the surround. Of 17 comparisons, the six discussed in the previous paragraph (for which outlines either reduced or removed the induced change) provided only weak support for the commonly held notion that outlines on maps counteract simultaneous contrast. Within this small sample, black outlines were less effective than white outlines, with only one instance of black removing induced change. Table 3.7 Each N=60 hex. size r g (6) gG (4.5) yY (4.5) b8 (6) l W (6) m K (4.5) rd R (11) gl 1: (4.5) bd Y (6) 82 dyx Comparison of Change Induced by Outlined Surrounds Relative to Contiguous and Control Surrounds white outlines black outlines induce: control: induce: control: -.17 +.23 " -.34 " +.03 +.52 " " -.87 " " (no data) +.11 -.95 " +.20 -.92 '” -.03 -.99 ” +.04 -.96 " -.03 -.80 " +.16 -.77 “t -.36 '” +.33 ‘ +.l3 +.80 "‘ +.57 " -.20 +.10 -.7O “ +.21 -.18 +.07 -.34 ” +.38 ” +.03 +.14 -.29 * Null Hypothesis for ’induce’ columns: The center test color with a contiguous inducing surround does not undergo a significantly greater shift in perception than with outlines separating the center and surround hexagons. Null Hypothesis for ’control’ columns: The center test color with outlined hexagons in the inducing surround does not undergo a significantly greater shift in perception than with the control surround. “ Reject H0 at .01 confidence level, one-tailed * Reject H0 at .05 confidence level, one-tailed Heterogeneous Surrounds Table 3.8 lists dyx for comparisons of heterogeneous-surround responses to the homogeneous-surround responses and to the control responses. Of the ten comparison pairs in Table 3.8, responses for five of the heterogeneous surrounds are significantly different from both responses for the corresponding homogeneous surround and for the control. These dyx are opposite in sign, indicating reduced changes in the center perception. Two (both 1 K combinations) of the heterogeneous surrounds induced the same amount of change as the corresponding homogeneous versions. Responses for three of the heterogeneous surrounds (m K, d K, rd R) were not significantly different from the control responses, indicating that breaking up the surround removed the induction effect. The results show that heterogeneous surrounds induce changes in perception. Consistent with the research literature, the induced change was most frequently intermediate between the effect of the homogeneous surround and the control surround. Results from this small sample of heterogeneous surrounds were not intended for prediction of the magnitude of these intermediate perceptions but to confirm that induction is not peculiar to the simple case of homogeneous surrounds. 84 Table 3.8 dyx Comparison of Change Induced by Heterogeneous Surrounds Relative to Homogeneous and Control Surrounds Each N=60 hex. homog. control size surround patchwork r R (11) +50 " * -.60 * * g G (4.5) +.72 " t -.87 " " yY (11) +.48 " -.38 " l W (6.5) +53 '* " -.39 " * l K (4.5) -.14 +51 " " 1K (11) -.09 +.77 * " m K (6.5) -.47 " " +.12 d K (4.5) -.41 " ‘ -.09 rd R (4.5) +.61 * * -.19 yl Y (4.5) +58 " " -.25 " Null Hypothesis for ’homogeneous surround’ column: The center test color with a homogeneous surround does not undergo a significantly greater induced shift in perception than with a heterogeneous surround of the same surround color and neutral. Null Hypothesis for ’control patchwork’ column: The center test color with a heterogeneous surround (composed of an inducing color and neutral) does not undergo a significantly greater induced shift in perception than with the control patchwork surround. " Reject H0 at .01 confidence level, one-tailed * Reject H0 at .05 confidence level, one-tailed Summary The analysis of results from Experiment 1 boils down to a simple set of broad guidelines for predicting changes in perception induced by surrounding colors. The rules build on the RG, YB, and I specifications of color appearance from the Hunt model. To predict lightness perceptions, use the J designations of lightness from the Hunt model: Calculate the difference between the lightnesses of the center and surround colors (center I - surround J). Mutiply the difference by 0.135 and add the product to the lightness of the center (if the surround is lighter than the center the difference is negative and therefore the center lightness is reduced). To predict hue perceptions, work on a graph showing the colors plotted in Hunt color space: Decide whether the surround hue is opposite or adjacent the center hue (they are adjacent if they are on the same side of neutral along an axis or are in the same quadrant). Draw a 90° buffer (a quadrant of a circle) extending from the center color toward the opponent complement of the surround color (extending equally to either side of this opponent direction). If the colors are adjacent, draw this buffer with a radius of 1.0 unit. If they are opposites, draw the buffer with a smaller radius of 0.5 units. (The 86 ’units’ are the same units in which the relative redness-greenness and relative yellowness-blueness axes of the graph are scaled.) Viewers will have difficulty distinguishing two colors if both of the following are true: The induction-affected lightness of the center color is within 12 lightness units of the perceived lightness of another color. The other color falls within the center’ 5 hue buffer or the other color’s buffer intersects that of the center color. This model for the prediction of induced rnisinterpretations of colors is intended for application to centers ranging in size from one-half to one degree of visual angle. The rules should hold whether colors are contiguous or finely outlined. Example applications of the model to specific map schemes are provided in Chapter 4, in which the success of predictions is analysed. CHAPTER 4 Experiment 2 : Methods, Results, and Discussion Experiment 2 tested predictions made by the model developed with Experiment 1. Subjects compared colors for three versions of each of 20 computer-displayed maps. They were asked to decide whether or not specific color pairs represented the same or different map categories. The comparison colors were presented with different inducing surrounds predicted to produce either incorrect responses or longer response times. These center colors were also compared with control surrounds of the same color. For the third version, map colors were adjusted such that the model predicted previously inducing surrounds would no longer cause an incorrect or slowed response. This chapter presents the test methods and results for the confirmatory second experiment. Discussion follows methods and results in a separate section. Methods Subjects As in Experiment 1, subjects were recruited with posters and newspaper advertising, tested individually, and paid ten dollars. Altogether, the responses of 30 subjects with normal color vision were obtained for Experiment 2. An equal number of males and females were tested and their ages ranged from 18 to 53 with a mean of 27. Seventeen subjects were non- students and 13 were students. The responses of an additional 5 subjects were 87 88 omitted from the final sample because of an interrupted test session, irregular response strategies, and drowsiness. Six paid pilot subjects were tested to clarify test instructions and establish a test time of approximately 50 minutes. Test Displays Figure 4.1 shows black and white versions of two test maps. The ten maps for Experiment 2 were produced with Map 11 (Pazner and others 1989). They were ”recoded,” ”combined," ”maximized,” ”overlayed," ”spread,” and otherwise manipulated variations on a set of seven geographic data layers for North St. Anne in Manitoba. These data were compiled in Map I] format by landscape architecture students under the supervision of Richard Perron at the University of Manitoba. The North St. Anne data were made available by Dr. Pazner. Square subsections of the Map 11 maps were imported to SuperCard, and titles and legends were then added. Thus, all test maps were generated from real data and included accurate titles and legends. Each map was presented with two different color schemes to produce 20 test maps. A variety of color scheme types were appropriate for the test maps. Qualitative schemes, for which nominally different categories were represented primarily with differences in hue, were used for eight maps (denoted by QL in Table 4.1). The two single-sequence (SS) schemes ranged from light to dark for single variables. Four are double-ended (DE) schemes for which darker colors of different hues emphasized both extremes of the distributions. Six bivariate schemes were included: two schemes combined quantitative (SS or DE) and qualitative (N for nominal: ’Ag.’ or ’Non-Ag.’ in this case) variables (SS/N and DE/N); two schemes combined qualitative categories with a nominal (’Inside’ or ’Outside’) variable (QL/N); and two schemes were two-variable quantitative schemes (58/ SS). Map3 Figure 4.1 Black-and-White Versions of Example Test Maps (reduced 48%) 90 Table 4.1 Experiment 2 Map Titles and Scheme Types Map Scheme Number Map Title and Color Scheme Number Type 9 Land Cover 1 QL 16 Land Cover 2 QL 11 Land Use 1 QL 2 Land Use 2 QL 18 Sand & Gravel Resources 1 QL 13 Sand & Gravel Resources 2 QL 6 Soil Series 1 QL 15 Soil Series 2 QL l4 Soil-Drainage Classes for Agricultural Land 2 SS 10 Soil-Drainage Classes for Forested Areas 2 SS 4 Soil-Drainage Classes for Agricultural Land 1 DE 1 Soil-Drainage Classes for Forested Areas 1 DE 19 Evaluation of Potential Campground Sites 1 DE 20 Evaluation of Potential Campground Sites 2 DE 7 Agricultural Land Uses and Aquifer Recharge 1 QL/N 8 Agricultural Land Uses and Aquifer Recharge 2 QL/N l7 Soil-Type Distributions for Agricultural versus Non-Agricultural Land Uses 1 SS/ N 3 Soil-Type Distributions for Agricultural versus Non-Agricultural Land Uses 2 DE/N 12 Soil Drainage and Distance to Surface Water 1 85/85 5 Soil Drainage and Distance to Surface Water 2 85/35 91 In all cases the map colors were chosen such that the perceptual organization of hue, value, and saturation provided a logical parallel to the organization of the mapped data. The schemes were designed by Eva Frank, a geography senior who had completed a Map Design course under my direction. Colors were not chosen with the purpose of encouraging induced misinterpretation or with restriction to particular positions relative to the axes of Hunt color space. A wide variety of colors were sought in the design of the set of maps. Once the schemes were completed, the luminance and chromaticity (L,x,y) of map colors were measured and converted to redness-greenness, yellowness-blueness, and lightness notations (RG,YB,J) using the Hunt model. I used the same parameter settings with the Hunt model as used in the first experiment (Chapter 2). The color schemes for each map were plotted in Hunt color space after the conversion to Hunt notation. One color pair predicted to fall victim to induced confusions was chosen from each of these graphs. To make these color-pair choices, I compared induced lightness changes and the overlap of hue buffers as recommended at the close of Chapter 3. For half of the test maps, two examples of the same color were predicted to look like representatives of different categories with surround induction. For the other ten maps, selected colors that were different were predicted to look the same with induction. Thus, ten color pairs for each case, Same or Different, were chosen from the graphs. Figure 4.2 provides a demonstration of the color selection process. The L x y, RG YB J, and R G B specifications for the colors of each of the 20 map schemes are listed in Appendix E. 92 Map 4: Soil-Drainage Classes for Agricultural land 1 Effect of adjacent dark-orange surround on gray-orange center: draw hue buffer 1.0 in blue-green direction; add 1 unit to gray-orange lightness for 72 J 72 = 71 + [0.135 x (center 71 - surround 61)] Effect of adjacent dark-purple surround on light-gray center: draw hue buffer 1.0 in green-yellow direction; add 3 units to light-gray lightness for 72 J 72 = 69 + [0.135 x (center 69 - surround 50)] Potential induced confusion is in overlap of induction buffers of gray orange and light gray, which both have lightnesses of 72. These two different colors may look the same. I adjusted by increasing the saturation of gray orange to prevent the buffers from overlapping. ,(1; 0 R ‘4 2 3 4 s s 3 ‘ 1 1 L 1 I 1 2 : ——640ran e‘1—"_S1 red_ -4 24 (It brown 8 . I > 1 . I 65 dk beige - 75 It beige I Id] 47 1* green 0 -1 78 gray__ _‘ ‘3 Graphs show all colors ‘ Wd :1: masters“: -2 __ . __. . 54 blue Numbers before names l I l are] values. Lines -3 connect surround and -—J affected colors. Effect of adjacent blue surround on dark-green center: draw hue buffer 1.0 in yellow-red direction; subtract 1 unit from dark-green lightness for 47 J 47 = 48 + [0.135 x (center 48 - surround 54)] Effect of opposite dark-brown surround on dark-green center: draw hue buffer 0.5 in blue-green direction; add 3 units to dark-green lightness for 51 l 51 = 48 + [0.135 x (center 48 - surround 24)] Map 13: Sand and Gravel Resources 2 Potential induced confusion is in overlap of dark-green induction buffer (tinted wedge) with blue-green that has a lightness of 61 (within 12 units of dark green’s induced lightness of 51). Thus, identical dark greens may look different because one surround shifts the color perception to overlap bl ue-green. I adjusted by increasing the greenness of dark green so its buffer no longer overlaps with blue green (open wedge). Figure 4.2 Example Color-Pair Selections from Experiment 2 Map Schemes 93 Three versions of each map were prepared once color confusions were identified: one had surrounds predicted to induce misinterpretation of color categories, one had matching control surrounds for both colors, and one had the scheme adjusted to remove induced difficulties. The test-color patches and surrounds were superimposed on the maps. This artifice was introduced in order to control the experiment more carefully than selection from the existing distributions would allow. The two colors that were compared on the three map versions were thus the exact same shape, size, and distance apart. The surrounds for induced, controlled, and adjusted cases were also equal in shape and size between maps. The color patches added to the maps were drawn with the characteristic stair-step edges imposed on the maps by the resolution of the original data. Arrows, of the same style used in Experiment 1, were positioned to point to the colors that subjects compared. Once the patches and arrows were positioned, the square maps were rotated or inverted differently for each of the map versions to reduce potential learning with repeated evaluations of colors in the same positions. Patches were positioned and rotations made such that the test colors were not located immediately adjacent to the map legends. Altogether, 150 color pairs were evaluated by each subject. The practice session at the beginning of the experiment comprised the ten base maps with different schemes than those of the test. Three versions for each of 20 color schemes were evaluated twice by each subject for 120 trials. In addition, one of each map with the adjusted scheme was included in the test with easily evaluated comparison colors. These comparisons were used to provide variety that could partly obscure the structure of the test and to provide more 94 presentations of the adjusted schemes. The test maps were presented in a unique random order to each subject. Test Procedure The first and second experiments were the same in overall structure. The second experiment was conducted using the same color monitor, curtained viewing area, and viewing distance used for Experiment 1. The same introductory protocol, background information questions, and color vision test were also used (Appendix A). For each map in the test, subjects were first presented with a gray screen showing two arrows and an OK button. They were instructed to be sure of the arrow positions and then click on OK using the mouse. The gray mask was then removed to reveal the map, legend, and two oblong buttons at the lower right labelled Same and Diff. The two arrows were pointing to two map colors and subjects were instructed to ”decide whether these two colors are from the same map category or different map categories.” During the practice session, the test administrator explained that the legend provided a key to the map colors and that each color represented a map category. These instructions were repeated for both a qualitative scheme with ten legend colors and a quantitative / qualitative scheme with a two-by-five legend arrangement. Subjects were also told that their responses were timed and that they should work quickly but accurately. The time between pressing OK and pressing Same or Diff. was recorded as the response time for each trial. The test instructions were detailed to ensure that subjects made comparisons of the map colors based on whether induced changes caused map colors to look like they belonged to a different category and not simply whether induction caused general differences in the colors. The latter effect 95 on map colors would not hinder a directed map reader as much as the first, more relevant, concern. The complete wording of the instructions is listed in Appendix A. Subjects were asked at the end of the test to describe how they made their decisions to check that they had followed these instructions. Once subjects clicked on the Same or Diff. button to make their response, the map was replaced with a homogeneous gray screen for approximately two seconds. The two arrows for the next map and OK were then superimposed on the gray, awaiting the click that initiated display of the next map. Display of the two arrows before each map allowed subjects to find them easily once they were overlayed on the map. This strategy was successful in removing search times (required to find the arrows) from the response-time data (Figure 4.3 pairs an arrow display with its map). The gray screen between maps was the same gray used between hexagonal displays in Experiment 1. Figure 4.3 Example Arrow Display and Corresponding Map The top display shows positions of test arrows before the induced version of Map 20 is displayed. The bottom display shows the arrows on the test map after the subject clicked on OK. (The map was displayed in full color.) 97 Hypotheses Times for subjects to decide whether colors represented the same or different map categories were analysed to answer the questions that follow: Does the model predict induced difficulty in color interpretation? CA Response times for the original colors with inducing surrounds (I) are greater than for adjusted colors with inducing surrounds (A). Does the adjustment remove the induced difficulty? A=C Response times for the adjusted colors with inducing surrounds (A) are the same as those for the original colors with control surrounds (C). The assumption underlying these hypotheses is that the longer a subject hesitated before making a response, the more difficult the decision, even though the response was correct. In the context of comparing map colors, this additional cognitive processing time may be a result of pausing to look at other map colors, looking at the legend, and looking back and forth to confirm a decision in addition to looking carefully at the colors in question. 98 Results Response Time Data Processing Tables 4.2 and 4.3 present results from Experiment 2. Table 4.2 lists the number of incorrect responses for each version of the map-color comparisons: induced (I), adjusted (A), and controlled (C). Recall that each subject saw each version of each map twice. If they answered correctly (Same or Diff.) on both trials then their two response times were averaged. If they answered correctly for only one of the two trials, this single response time was used in the analysis. Incorrect responses on both trials for any of the three versions resulted in removing that subject from the analysis of response times for that particular map scheme. The second column (N) of Table 4.3 lists the number of response-time triplets (t, A, and C times for one subject) that were used for analysis of each map scheme. The samples ranged from all 30 to a minimum of 13 subjects for the response-time analyses. A missing response time (both trials incorrect) required omission of the subject because the experiment was structured for the sensitive one-way repeated-measures analysis of variance (AN OVA). The repeated-measures approach removes the variability between subjects from the analysis of group differences in mean response times (differences between subjects was expected but would be relevant to discussion of a question separate from generalization about the model performance). The I, A, and C versions of the maps were the fixed effect under study. Phrasing equivalent to a repeated measures design for one-way ANOVA is a randomized block design (Wilkinson 1989 p. 219) with subjects as blocks responding to the maps in a random order. 99 Table 4.2 Experiment 2 Maps and Number of Incorrect Responses Each N=6O Responses Map Incorrect Number Map Title and Color Scheme Number I A C 1 Soil-Drainage Classes for Forested Areas 1 18 3 6 2 Land Use 2 3 1 2 3 Soil-Type Distributions for Agricultural versus Non-Agricultural Land Uses 2 18 2 3 4 Soil-Drainage Classes for Agricultural Land 1 3 0 0 5 Soil Drainage and Distance to Surface Water 2 7 6 2 6 Soil Series 1 5 0 0 7 Agricultural Land Uses and Aquifer Recharge 1 30 3 2 8 Agricultural Land Uses and Aquifer Recharge 2 3 2 2 9 Land Cover 1 29 8 10 10 Soil-Drainage Classes for Forested Areas 2 5 4 11 Land Use 1 2 1 12 Soil Drainage and Distance to Surface Water 1 13 14 0 13 Sand 8: Gravel Resources 2 24 11 O 14 Soil-Drainage Classes for Agricultural Land 2 26 22 0 15 Soil Series 2 26 27 0 16 Land Cover 2 3 4 1 17 Soil-Type Distributions for Agricultural versus Non-Agricultural Land Uses 1 24 28 3 18 Sand 8: Gravel Resources 1 21 22 1 19 Evaluation of Potential Campground Sites 1 37 22 1 20 Evaluation of Potential Campground Sites 2 21 19 0 Table 4.3 100 Experiment 2 AN OVA of Logged Response Times The probability (p) values shown in bold are those small enough that the null version of the associated hypothesis may be rejected. The hypotheses are listed in the previous section of this chapter. N subi Map 1 24 2 30 3 23 4+ 29 5+ 26 6 29 7 20 8 29 9 19 10+ 28 11” 30 12 26 13 21 14 20 15+“ 16 16 29 17+ 15 18 21 19 13 20+ 20 all P IAC 1.21 22.37 9.01 12.55 11.19 10.52 10.25 5.68 5.00 8.23 5.76 13.83 15.40 8.87 10.06 17.69 14.83 8.95 11.35 14.30 F IA 1.99 44.65 15.09 20.86 11.67 19.96 19.66 8.39 8.07 4.86 4.15 5.86 3.05 0.31 0.40 0.33 0.08 0.07 0.00 4.10 1 df contrasts F F A C CI 1.61 0.02 9.53 12.92 0.21 11.73 0.25 16.52 1.27 20.63 1.78 9.82 2.03 9.07 0.00 8.64 0.05 6.87 3.41 16.42 1.77 11.35 8.03 27.61 13.61 29.55 11.12 15.17 17.32 12.47 23.41 29.34 20.91 23.51 14.37 12.41 17.06 16.98 28.08 10.72 .017 P CI .887 .001 .001 .000 .000 .003 .005 .005 .013 .000 .001 .000 .000 .000 .001 .000 .000 .001 .000 critical probability values: .050 .017 .017 (Bonferroni levels) P P P I A C I A A C .309 .165 .211 .000 .000 .003 .001 .000 .648 .000 .000 .617 .000 .001 .265 .000 .000 .187 .000 .000 .163 .006 .005 .966 .012 .007 .827 .001 .032 .070 .005 .046 .188 .000 .019 .007 .000 .088 .001 .001 .578 .002 .000 .533 .000 .000 .565 .000 .000 .785 .000 .001 .790 .000 .000 .992 .000 .000 .050 .000 .002 " removed subject with extreme residual to improve normality or variance *1 two response-time groups not normal (gap in histogram) ”1 variances not equal with natural logarithm (LN) transformation 101 Fulfillment of ANOVA Assumptions The assumptions of AN OVA required that groups of response times for each map version were normally distributed with equal variance. The response-time groups displayed an expected skew toward shorter times with a tail of subjects taking time to think before making their decisions. likewise, variances were greater with the longer response times for the more difficult comparisons. Of the 60 responsetime groups (I, A, and C groups for 20 maps), 23 were not normal (the nonparametric Lilliefors test was used to evaluate normality). Heteroscedasticity plagued the response variances for all but four of the maps (Bartlett’s test was used to compare variances). Two techniques were used to transform these groups of response times to normality and homoscedasticity. The natural logarithms of response times for all maps were used for the ANOVA. For five maps, omission of single subjects with extreme residuals was necessary to meet the assumptions (these are marked 1- in Table 4.3). The log transformation and the removal of outliers produced groups that were all normally distributed with the exception of the adjusted and control groups for Map 11 (H in Table 4.3). Examination of these histograms revealed a gap in the frequency distributions that “was resistant to transformation to normality. In addition, differences in response variance for Map 15 did not yield to removal of obvious outliers or alternative transformations. Wilcoxan signed-rank testing, a nonparametric treatment of the data, was used to confirm the ANOVA results for all of the maps with a particular eye toward discrepancies with the results for Maps 11 and 15. The Wilcoxan test was used to compare pairs of response-time groups (I to A, A to C, and C to I). The test required calculation of differences in respOnse times for each subject 102 and ranking the absolute values of the differences. The sum of ranks for positive and for negative time differences were compared by calculating a probability that there is no difference between the two groups of matched times. Overall, the Wilcoxan test produced only three, out of 60, hypothesis evaluations that were different than the AN OVA results. None of these discrepancies occurred for the results for Maps 11 or 15. Because AN OVA is considered relatively robust and because of agreement with the Wilcoxan results, the ANOVA results for these two maps that do not meet the normality and homoscedasticity assumptions are reported along with results for the other maps in Table 4.3. The specifics of the supporting Wilcoxan analysis are not listed, though the three discrepancies with the ANOVA are noted in Figure 4.4 that appears in the next section. ANOVA Results In Table 4.3, the F ratios and associated probabilities resulting from the ANOVA are reported. The F statistic indicates the ratio of the variance of mean response times between groups to the variance within groups. The probability (p) values show the likelihood that this difference occurred by chance. For the overall comparisons of the three groups of times for each map, a probability of more than .05 occurred for only one map (listed as Map 1 in Table 4.3). The hypotheses listed in the previous section detail a priori contrasts planned between the I, A, and C groups of response times. These one-degree- of-freedom contrasts are made within the ANOVA procedure (Dowdy and Wearden 1983 p. 278, Wilkinson 1989 p. 197-201). The results of the contrasts between each pair of response-time groups are listed in Table 4.3. The 103 Bonferroni procedure was used to establish a critical probability level of .017 for each contrast. This level is one third of the overall critical value of .05. The .017 level was used to maintain overall protection against a Type 1 error at .05 for the three interdependent contrasts (Wilkinson 1989 p. 201). The contrasts for which the associated null hypothesis may be rejected are shown in bold type in Table 4.3. The maps in Table 4.3 are arranged in a systematic order. The first map was the single example for which there was not a significant difference between the I and C response-time means. The second group of eight maps (2 through 9) were those for which the I means were significantly slower than the A means. The adjusted-scheme mean for Map 2 was also significantly faster than the control, but the remaining seven A and C means were not significantly different. For Maps 10 and 11 there was no significant difference between either I and A or A and C means. This lack of difference indicated that the adjusted group means were between those of I and C, which were close but still significantly different (C < I). The last set of maps (12 to 20) were those for which the adjustments did not produce a significant difference between I and A mean times, with I and A both significantly slower than C. Graphic Data Summary Figure 4.4 provides a graphic summary of the response time data. Tukey box plots were constructed showing the three distributions for each map. Within each of the 20 frames, the left box shows the logged response times for the induction-affected colors (I), the middle box shows the A response times, and the right box shows the C times (I, A, and C labels appear below the boxes). The bold line across each box marks the median of the group times and the box encloses the interquartile range. The top and bottom edges of the 104 box are referred to as upper and lower ’hinges.’ The maximum range encompassed by the box whiskers is four times the interquartile range (1.5 of the interquartile range above and below the hinges), but the whiskers extend only to the last data value within this range. Response times beyond the whiskers are plotted with asterisks. The box plots provide a nonparametric view of the logged data that complements the parametric ANOVA results, which are summarized at the bottom of each frame. The boxes are also shaded to echo the AN OVA results. For boxes with like shading, there was no significant difference between the means of the response times (I = A, A = C, C = I below boxes). Note that the medians were close when the means were not significantly different. Likewise when I > A, A > C, or C <1, the boxes have different shadings and the differences between medians support the ANOVA results. The instances in which the Wilcoxan analysis produced differing evaluations of hypotheses are indicated in parentheses above the ANOVA summary line (Maps 9, 12, and 20). 105 I III NU.“ ResporBe-T' ° Explanation of Tukey Box Plots "m Groups. dam values Induced Controlled . outside range 8 Adjusted ‘ ofwhiskers E l: 3 2 -upperhinge o 8 . interquartile median 8 § . range a? . -|ower hinge 3 " 1 —whisker x ‘ last data value — Diff: withinl.50f SummaryofANOVA Results——> I = A = C = l interquartile 0 range from hinge Map 1 3 g 2- r .. - m . E I - 2 n - E. 1. W m < , . _ w fl . m m oDiftzl>AA=CA=CA=CA=CA=CA=CA=C) oSarnezl=A>C< Sarne:|=A>CC< Map12 Map13 Map" 3 . if“ ~ - + z m ‘ ' -'1- m - oSame:I=A>C< Samezl=A>CC< Map15 Map16 Map17 3 32’ ’ l i 1 * .. m H m 51. m . m - m (<) oSame:l=A>C< SamezlsA>CC_ ‘ to test shift away from 1 ' 29 brown the surround (tinted . l 35 dksurround buffer), rather than 0 .mgenl'; toward the opponent ? ad] 40 mfi'mgwra complement of the . surround ' surround (’l’ buffer) 46 and l a _1 62 GNY 55 ._ magen a ‘ 55 8"Y' -2 -—green-blue "“39““ ‘ medium magenta '3 is adjusted darker '— ‘ _ to be out of range 4 A 7 dk b "e of gray-magenta -— . lightness buffer 1 l I -5 I I I I I l V I I I I -2 -1 o 1 2 3 4 5 6 G R Map 8: Agricultural Land Uses and Aquifer Recharge 2 4 Comparison constructed to test shift toward the 3 opponent complement dk of the surround (tinted 2 64 dk buffer), which is more sum ll toward the surround >- 1 79 " {9"0w'8'ee" ,5; 35 j, color than the ’1’ buffer 78 green ' 78 pink I 0 61 dk gray l a -1 -2 Figure 4.5 Comparison of Desaturated Surround Predictions for Maps 1 and 8 111 Summary My overall conclusion from the Experiment 2 results is that it is easier to induce matched colors to look different than to induce different colors to look matched. Thus, colors that are already different can be adjusted more readily to repair induced ambiguity. The model accurately predicts these adjustments. When colors are the same and appear with different surrounds it is more difficult to confidently decide that they match. This study was not adequately controlled to be sure of the nature of the failure in the model adjustments for Same responses; an additional control presenting center colors with different surrounds that should not induce confusion with other map colors was needed (controls I used always had identical surrounds for color pairs). CHAPTERS Conclusions and Recommendations I have developed a set of guidelines for the prediction of surround- induced changes in color appearance. The specific rules for applying the model to a set of map colors are detailed at the close of Chapter 3. They provide an extension of the Hunt model of color appearance to predict simultaneous contrast effects between colors on complex computer-displayed maps. The model performed well in subsequent testing with a set of 20 maps, although results for Same and Different responses differed. In the course of developing the model and testing it, additional research questions came to light. I will briefly summarize results and discuss recommendations for further work in this final chapter. Lightness Induction Darker surrounds induce perceived lightening of a color and lighter surrounds induce darkening. To predict the effect, the contrast in lightness between a center color and its surround is calculated. A small proportion of this contrast is then added to the color of the center to estimate its perceived lightness. The greater the contrast between center and surround, the greater the lightness shift. Lightness perceptions were, however, less affected by induction than I had expected. This may be partly because the computer monitor is not capable of 112 113 creating a great range of lightnesses with which to work. The small shifts in lightness may also be due to testing simultaneous contrast effects in the context of complex displays rather than with isolated center-surround stimuli. The issue of recognizing the interplay of lightness versus brightness evaluations and contrast versus constancy effects was discussed in Chapter 1. Lightness induction was a significant effect on map color perception in this research, and I have modelled it in a manner applicable to map-reading situations. Chromaticity Induction The shift in hue and saturation of a color is estimated by constructing a buffer, or zone of potential perceptions, that extends from the color toward the opponent complement of the surrounding hue. If the center and surround are similar in hue, the buffer extends a greater distance than if the center and surround are dissimilar in hue. Yellow surrounds cause a color to shift toward blue and blues cause a shift toward yellow. Red surrounds cause a shift toward green and greens cause a shift toward red. These directions of shift are applied simultaneously for hues that do not fall on these opponent axes (for example, a green-blue causes a shift toward red-yellow). Chromaticity shifts that occur toward neutral are induced desaturations of color, rather than perceived hue changes. Saturation Importance Shifts in saturation induced by like hues were the largest shifts witnessed in this research. Saturation is generally a less well understood color attribute and is infrequently used as a primary contrast variable in the construction of 114 map schemes. A thematic scheme that makes use of a progression from a saturated hue to neutral (white, gray, or black) does, however, require systematic saturation differences to produce an orderly scheme, though lightness changes or hue may define the look of the map. In teaching I have found that novice color users will describe the differences in hue and lightness between colors but will not describe differences along the third dimension of saturation (though they would not say that, for example, orange and brown or green and olive match). This often unrecognized variable is apparently particularly susceptible to induced shifts. This unexpected aspect of perception suggests that guidelines to pinpoint specific color interpretation difficulties would aid design because many people will not overtly identify a saturation difference, making it difficult for them to notice that the saturation difference is compromised by induction. This leads me to wonder whether the reason this dimension is readily affected by induction is because people pay less overt attention to it. The result that surrounds of like hue produced greater shifts than more distant hues was also unexpected. The general interpretation of opponent effects with which I began suggested that greater contrasts along a red-green or yellow-blue axis would produce greater shifts, as would greater contrast along the lightness axis. The sampling of color space used in the first experiment did not include a sufficient range of color combinations with which to fully address this aspect of the model (more center-surround contrasts with like hues were needed). Recommendations for Saturation Research Additional research would reveal whether increasing saturation contrasts within like-hue combinations produce increasing or decreasing perceived 115 shifts. The reduction of shifts with greater difference may be restricted to the qualitative differences I tested (interaction between different hues, and hue shifts with differences in lightness). A subjective interpretation of what I saw on the displays was that centers with surrounds that were very different in color were affected by induction but the interaction was so obvious that it could be compensated by subjects in their analytical evaluation of the colors. More work could also be done to determine how desaturated and gray surrounds affect more saturated centers, as discussed in previous chapters. Is the shift still toward the surround’s complement or away from the surround toward increased saturation? Model Applications The model is readily implemented for evaluation of a set of map colors that are specified in Hunt units of lightness (I) and relative RG and YB. Potential induced confusions between map colors are found by applying the model to each combination of map colors related as center and surround. Given the set of colors on a map, induction may cause colors to have overlapping hue perception buffers and lightness buffers. These colors will be difficult to distinguish on the map when they appear with the offending surrounds that cause the shifts. This model provides an objective check of a set of colors to warn of potential difficulties for map readers interpreting the mapped data. The logic of a scheme may also be considered at the stage of evaluating induced shifts within a map scheme. If the potential confusions occur with center-surround combinations that would not appear on the map, the induction relationship is not of concern for that map scheme. For example, 116 the effect of a high—elevation color on a low-elevation color may seem to be a problem, but the logic of the mapped terrain will preclude their occurrence in a contiguous center-surround relationship. An alternative way to go about the evaluation of a scheme would be to check all center-surround combinations that do appear on the map. Three different sizes of center colors, however, did not produce significant differences in induction magnitudes in this research. This lack of difference casts doubt on the potential advantage of prediction models sensitive to size and spatial relationships between map polygons. Because maps frequently present a limited set of colors, the comparison of all color combinations provides a more comprehensive evaluation of the scheme and requires less calculation. Likewise, heterogeneous surrounds will produce the same Or a reduced shift in appearance, and in these cases looking at the effects of all potential combinations of centers and surrounds provides a cautious but not inappropriate warning. Recommendations for Applications Research Investigating a greater range in color sizes is an obvious extension of this work. The model I have developed is suitable for centers ranging from one- half to one degree visual angle (within this range I found no systematic differences in susceptibility to induction). The surrounds in Experiment 1 were of a uniform large size. In comparison, the inducing surrounds on the maps in the second experiment were quite small in many cases and surround effects were still evident in the response-time analysis. Additional research could clarify the effects of the surround-size parameter on the induction relationship. 117 Color difference is a companion of induced change in the discrimination of colors. I have not evaluated the importance of the initial differences between map colors in my analysis of the Experiment 2 schemes, relying on careful design to meet that requirement. Basic color-difference guidelines (least practical differences; not just noticeable differences) would be an appropriate starting point in map color selection for the novice and detecting induced shifts that compromise these differences would be added assistance. The calculation of induction magnitudes may be used to establish practical color differences for map design. Building on the Hunt model may limit use of the guidelines I have set forth. Because the final induction model provides fairly broad recommendations, the advantages of using the Hunt specifications may be outweighed by the advantages of using a simpler perceptually-scaled color system such as CIELUV. The formulae for CIELUV are more readily programmed and they are invertible, which would aid an extended model’s ability to specify adjustments to colors to remove induced confusions. Testing a simpler color system against the advantages of accounting for the effect of adaptation on color appearance, for example, would be a useful extension of the research. Final Comments An underlying assumption of highly-focussed map design research, such as this induction work, is that cartographers bear responsibility for efficiency in the communication of mapped information. Map readers, however, share this responsibility for success of the two-way communication process. The map reading tasks used in this research were realistic but required only 118 simple color comparisons. The overall perceptual structuring of a map scheme, in addition to the differentiability of colors, should have important effects on the success of map readers motivated to solve more complex spatial problems. The apparent difficulty of deciphering a map color scheme may also affect the perceived usefulness of a map and may affect the amount of effort that a map reader is willing to invest in using the map. Research concerned with the broader issue of color scheme effects on the understanding of spatial patterns is needed. In developing a model of induction, I have drawn on a large body of color theory. The induction model has a deceptively simple structure because it is based on a comprehensive appearance model which in turn is based on decades of color perception research. Bringing this theory into the applied realm of map color perception offered the challenge of deciding which aspects of the experiments could be controlled and which could vary without corrupting the results. The model is intended to be applied to color combinations from throughout color space, which is an expansive goal. The perception-buffer approach for the prediction of both lightness and chromaticity induction was chosen to accommodate 90 percent of map readers, rather than limiting conclusions to perceptions of an artificial average reader. The goal of my guidelines for predicting induction is to provide a practical tool for the design of successful color schemes for computer-displayed maps. LIST OF REFERENCES LIST OF REFERENCES Agoston, GA. 1987. Color Theory and Its Application in Art and Design. 2nd ed. New York: Springer-Verlag. Agresti, A. 1981. Measures of Nominal-Ordinal Association. journal of the American Statistical Association 76(375):524-529. Appleton, B. 1990. SuperCard 1.5. San Diego: Silicon Beach Software Inc. Arend, L.E., and R. Goldstein. 1987. Simultaneous Constancy, Lightness, and Brightness. Journal of the Optical Society of America A 4(12):2281-2285. Bartleson, CJ. 1979. Changes in Color Appearance with Variations in Chromatic Adaptation. Color Research and Application 4(3):119-138. Blackwell, K.T., and G. Buchsbaum. 1988. The Effect of Spatial and Chromatic Parameters on Chromatic Induction. Color Research and Application 13(3):166-173. Bohrnstedt, G.W., and D. Knoke. 1982. Statistics for Social Data Analysis. Itasca, Illinois: F.E. Peacock. Breneman, EJ. 1977. Perceived Saturation in Complex Stimuli Viewed in Light and Dark Surrounds. Journal of the Optical Society of America 67(5):657-662. Brewer, C.A. 1989-90. Color Chart Use in Map Design. Cartographic Perspectives 4:3-10. Brown, A. 1982. A New ITC Colour Chart Based on the Ostwald Colour System. The ITC lournal 2:109-118. Campbell, I. 1991. Introductory Cartography. 2nd ed. Dubuque, Iowa: Wm. C. Brown. 119 120 Castner, H.W. 1980. Printed Color Charts: Some Thoughts on Their Construction and Use in Map Design. ACSM-ASP Technical Papers, 40th Annual Meeting:370-378. Chevreul, M.E. 1839/1981. The Principles of Harmony and Contrast of Colors and Their Applications to the Arts (Introduction and Notes by F. Birren). Translated by C. Martel in 1854 from 1839 French edition. New York: Van Nostrand Reinhold. Cox, CW. 1980. The Effects of Background on the Equal Value Gray Scale. Cartographica 17(1):53-71. De Valois, R.L., M.A. Webster, K.K. De Valois, and B. Lingelbach. 1986. Temporal Properties of Brightness and Color Induction. Vision Research 26(6):887-897. Dent, B.D. 1990. Cartography: Thematic Map Design. 2nd ed. Dubuque, Iowa: Wm. C. Brown. Dowdy, S., and S. Wearden. 1983. Statistics for Research. New York: John Wiley and Sons. Eastman, R.J. 1986. Opponent Process Theory and Syntax for Qualitative Relationships in Quantitative Series. The American Cartographer 13(4):324—333. Eichengreen, J.M. 1976. Unique Hue Loci: Induced Shifts with Complementary Surrounds. Vision Research 16:199-203. Ejima, Y., and S. Takahashi. 1983. Relationship Between Chromatic Induction and Spatial Variables: An Integrated Explanation in Terms of Element-Contribution Function. Journal of the Optical Society of America 73(2):203-207. Evans, R.M. 1948. An Introduction to Color. New York: John Wiley & Sons. Fach, C., and LT. Sharpe. 1986. Assirnilative Hue Shifts in Color Gratings Depend on Bar Width. Perception and Psychophysics 40(6):412-418. Gilchrist, A.L. 1988. Lightness Contrast and Failures of Constancy: A Common Explanation. Perception and Psychophysics 43(5):415-424. Guth, S.L., R.W. Massof, and T. Benzschawel. 1980. Vector Model for Normal and Dichromatic Color Vision. Journal of the Optical Society of America 70(2):]97-212. 121 Heinemann, EC. 1972. Simultaneous Brightness Induction. In Visual Psychophysics. pp. 146-169. Edited by D. Jameson and L. M. Hurvich. New York: Springer-Verlag. Hunt, R.W.G. 1982. A Model of Colour Vision for Predicting Colour Appearance. Color Research and Application 7(2, pt 1):95-112. Hunt, R.W.G. 1985. Perceptual Factors Affecting Colour Order Systems. Color Research and Application 10(1):12-19. Hunt, R.W.G. 1987a. The Reproduction of Colour in Photography, Printing and Television. 4th ed. Tolworth, England: Fountain Press. Hunt, R.W.G. 1987b. Measuring Colour. Chichester, England: Ellis Horwood. Hurvich, L.M. 1981. Color Vision. Sunderland, Massachusetts: Sinauer Associates. Imhof, E. 1982. Cartographic Relief Representation. Translated by H. J. Steward. New York: Walter de Gruyter. Jacobsen, A., and A. Gilchrist. 1988. Hess and Pretori Revisited: Resolution of Some Old Contradictions. Perception and Psychophysics 43(1):7-14. Jameson, D., and L.M. Hurvich. 1964. Theory of Brightness and Color Contrast in Human Vision. Vision Research 4:135—154. Keates, LS. 1982. Understanding Maps. New York: Longman Group. Kimerling, A.J. 1975. A Cartographic Study of Equal Value Gray Scales for Use with Screened Gray Areas. The American Cartographer 2(2):119-127. Kimerling, A.J. 1985. The Comparison of Equal-Value Gray Scales. The American Cartographer 12(2):132-142. Kinney, J.A.S. 1962. Factors Affecting Induced Color. Vision Research 2:503- 525. Krauskopf, J., Q. Zaidi, and M.B. Mandler. 1986. Mechanisms of Simultaneous Color Induction. Journal of the Optical Society of America A 3(10):1752-1757. Land, EH. 1986. Recent Advances in Retinex Theory. Vision Research 26(1):7-21. 122 Leibowitz, H., F.A. Mote, and W.R. Thurlow. 1953. Simultaneous Contrast as a Function of Separation Between Test and Inducing Fields. Journal of Experimental Psychology 46(6):453-456. Livingstone, M., and D. Hubel. 1988. Segregation of Form, Color, Movement, and Depth: Anatomy, Physiology, and Perception. Science 240:740-749. Marsden, A.M. 1969. An Elemental Theory of Induction. Vision Research 9:653-663. Moellering, H., and A.J. Kimerling. 1990. A New Digital SlopeAspect Display Process. Cartography and Geographic Information Systems 17(2):]51-159. Monmonier, M.S. 1980. The Hopeless Pursuit of Purification in Cartographic Communication: A Comparison of Graphic-Arts and Perceptual Distortions of Graytone Symbols. Cartographica 17(1):24-39. Muller, J.C. 1979. Perception of Continuously Shaded Maps. Annals of the Association of American Geographers 69(2):240-249. Pazner, M., K.C. Kirby, and N. Thies. 1989. Map 11 Map Processor: A Geographic Information System for the Macintosh. New York: John Wiley and Sons. Peterson, M.P. 1979. An Evaluation of Unclassed Crossed-Line Choropleth Mapping. The American Cartographer 6(1):21-37. Pitt, I.T., and L.M. Winter. 1974. Effect of Surround on Perceived Saturation. Journal of the Optical Society of America 64(10):1328-1331. Robertson, P.K., and J.F. O'Callaghan. 1986. The Generation of Color Sequences for Univariate and Bivariate Mapping. IEEE Computer Graphics and Applications 6(2):24-32. Robinson, A.H. 1952. The Look of Maps: An Examination of Cartographic Design. Madison: The University of Wisconsin Press. Robinson, A.H. 1967. Psychological Aspects of Color in Cartography. International Yearbook of Cartography 7:50-61. Robinson, A.H., R.D. Sale, J.L. Morrison, and RC. Muehrcke. 1984. Elements of Cartography. 5th ed. New York: John Wiley 8: Sons. 123 Shevell, S.K., and M.F. Wesner. 1989. Color Appearance Under Conditions of Chromatic Adaptation and Contrast. Color Research and Application 14(6):309-317. Somers, RH. 1962. A New Asymmetric Measure of Association for Ordinal Variables. American Sociological Review 27(6):799-811. Somers, R.I-I. 1980. Simple Approximations to Null Sampling Variances. Sociological Methods and Research 9(1):115-126. Spiker, A., S.P. Rogers, and J. Cicinelli. 1986. Selecting Colour Codes for a Computer-Generated Topographic Map Based on Perception Experiments and Functional Requirements. Ergonomics 29(11):1313-1328. Takahashi, S., and Y. Ejima. 1983. Chromatic Induction as a Function of Wavelength of Inducing Stimulus. Journal of the Optical Society of America 73(2):190-197. Takasaki, H. 1969. Von Kries Coefficient Law Applied to Subjective Color Change Induced by Background Color. Journal of the Optical Society of America 59(10):]370-1376. Troscianko, T.S. 1977. Effect of Subtense and Surround Luminance on the Perception of a Coloured Field. Color Research and Application 2(4):153- 159. Troscianko, T.S. 1982. Saturation as a Function of Test-Field Size and Surround Luminance. Color Research and Application 7(2, pt 1):89-94. Valberg, A. 1974. Color Induction: Dependence on Luminance, Purity, and Dominant or Complementary Wavelength of Inducing Stimuli. Journal of the Optical Society of America 64(11):1531-1540. Ware, C. 1988. Color Sequences for Univariate Maps: Theory, Experiments, and Principles. IEEE Computer Graphics and Applications 8(5):41-49. Ware, C., and W.B. Cowan. 1982. Changes in Perceived Color Due to Chromatic Interactions. Vision Research 22:1353-1362. Whipple, W.R., H. Wallach, and E]. Marshall. 1988. The Effect of Area, Separation, and Dichoptic Presentation on the Perception of Achromatic Color. Perception and Psychophysics 43(4):367-372. Wilkinson, L. 1989. SYSTAT: The System for Statistics. Evanston, Illinois: SYSTAT Inc. 124 Wood, M. 1968. Visual Perception and Map Design. The Cartographic Journal 5(1):54-64. Yund, E.W., and J.C. Armington. 1975. Color and Brightness Contrast Effects as a Function of Spatial Variables. Vision Research 15:917-929. APPENDICES APPENDIX A APPENDIX A Experiment Procedures The procedures for Experiments 1 and 2 are described in detail in this appendix. Much of the protocol is common to both experiments, and the shared procedures are designated BOTH in the script that follows. Those instructions unique to the individual experiments are headed by EXP 1 or EXP 2 labels. The four experiment administrators adjusted their phrasing for individual speaking styles, but they were trained with these scripts. All four strove for accuracy and consistency in presentations to subjects. BOTH Room Preparation: Computer on (half-hour warm up) Chair in hallway (if multiple appointments) Answering machine to "announce" and volume to "minimum" (EXP 1) Maps shuffled (each Exp. 2 and first half of Exp. 1) Lights on Fan in position and on appropriate speed Black curtain up Chin rest in place, with napkin cover Window blinds open Prepare clipboard and pencil consent form on top folded background form with vision test up subject number on background form $10 in pocket 125 126 Introduction: Welcome subject Flip "experiment in progress" sign EXP 1 You are participating in research on color perception. We are studying perception to improve computer-map design. I'll first test your color vision. Then you'll look at a series of colors on the computer screen and select colors similar in appearance. EXP 2 You are participating in research on the perception of map colors. We are studying perception to improve computer-map design. I‘ll first test your color vision. Then you'll look at a series of maps on the computer screen and decide whether or not specific map colors match. BOTH This is a black curtain that will give us controlled surroundings for the computer screen. Before we start, could you please sign this consent form. It‘s required by the university for experiments with people. Hand clipboard and pencil to subject BOTH Color Vision Test: Now we'll do the color vision test. Turn to the next sheet on your clipboard and fill out the questions on the left half of the page. Turn off both overhead lights Wait until questions completed OK, please stand facing me behind the line over by the window. That gives us good daylight for the test. This is a standard test that you may have seen before. It will go quickly. I'll show you sets of dots and you tell me what numbers you see in them. Open Ishihara book Stand behind other line (75 cm) Hold plates perpendicular to subject's line of sight with good lighting 127 As you look at the pages, say out-loud the numbers you see and also write them down in the right-hand column on the form you have. If you don't see a number on a page, tell me so and put an 'X' on the form. Go through five plates. They should answer in 3 seconds - quickly. Note hesitations on form. When complete, put away plates and take forms. Write down 'normal' or ’deficient' at bottom of form. You have normal color vision. Let's go to the experiment now. If deficient: According to this test, you have anomalous color vision... don 't dwell on a diagnosis BOTH Preparation: Lower black curtain Hook string from back of curtain to bookcase Turn on both overhead lights I have a chin rest here to position your eyes at a constant distance from the screen. Let's adjust the chair so this will be comfortable for you. You don't need to prop your head on the chin rest - just position your head so your chin is at the bar. Adjust chair, tighten well Is this fan speed good for you? It takes about 45 minutes to go through this test, so it will get stuffy in here. Feel free to adjust the fan or the chair later on. Adjust fan Have you used a mouse before? Are you left or right handed? How's that? Adjust mouse pad, center mouse Close black curtains 128 EXP 1 Practice Session: I first have a series of 10 practice images. As we work through these, you can get used to using the mouse and can ask me any questions you may have. Press 'Start Practice' button Each display will look something like this. There's a center color here with an arrow to it (point). Around the outside there are 6 choices, also with arrows (paint). Choose one of these 6 choices (point again) that looks most similar in appearance to this center colorfpoint again). If familiar with mouse: To make you're selection, click on the chosen color. The computer will record your choice. If not: You move the mouse on this pad with your hand and the cursor (the little hand on the screen) will move with it. To make your selection, move the hand over to the color you've chosen and press this button on the mouse (demonstrate). The computer then records your choice. You'll get good at this with just a little practice. Click to get gray between displays When you click the mouse, the colors will disappear as they just did. This gray screen will appear between cards. The computer takes some time to put up the displays, since they are large graphics files. It also gives you a chance to rest. Do a few of these screens to practice. 129 EXP 1 Practice (continued) Interrupt at about third display or when appropriate Be sure to look at all 6 before making your choice, but don't take too long if you're having trouble choosing. Just choose the one that looks closest in appearance to the center color. Try to go through the displays quickly. Break on fifth Every twenty-five cards during the real test, this break message will come up. This give you a chance to rest your eyes, shift your position, turn your neck, and such. I'll also ask you a question for our background information. I'll ask you the first background question now: (Q 1 on form) When you are ready to continue with the test, click on that message on the screen. Do you have any questions? Continue to end of practice Do you want to do any more practice before we start? If not, begin experiment. 130 EXP 2 Practice Session: I first have a series of 10 practice maps. As we work through these, you can get used to using the mouse and can ask me any questions you may have. Press 'Start Practice' button Before each map, this gray screen is displayed. On it you'll see two arrows. They will be pointing to two map colors. This little hand is the cursor you control with the mouse (demonstrate if not familiar with mouse). When you are sure of the positions of the arrows, move the cursor to this OK button and click the mouse key. Click OK to display map Here is an example map. The arrows point to these two colors (poi nt). I want you to decide whether the two colors are from the same map category or from different categories. The legend for the map, over here, provides a key to all of the map colors. On this map there are 10 different map categories. These two colors look like they are examples from the same yellow category. One may look a bit lighter or otherwise different than the other, but since there is only one yellow in the legend you would choose Same. To make your response, move the cursor to the Same button and click. If the colors were from different categories, you would click on the Diff. button. Click to get gray between maps When you click the mouse, the colors will disappear as they just did. This gray screen will appear and then the next set of arrows will come up. Press OK when you are ready. After you press OK, leave the mouse still so you are right between the two choices. Move to your choice once you make the decision. 131 EXP 2 Practice (continued) This map also has 10 different categories represented by 10 colors. Though these two map colors are similar, there are two yellow-brown colors in the legend, so you would decide that the arrows point to colors from different categories. If there was only one yellow-brown in this legend, you would decide that the colors were from the Same category, despite their differences. Subject makes choice. When you press the Same or Diff. button, the computer records your choice. It also records the amount of time you took to respond; the time between pressing OK and Same or Diff. I want you to answer as quickly as you can while still being accurate. Don't race through, making sloppy choices. Do click the button, however, as soon as you decide whether the map categories for the colors are the same or different. Continue with these maps to practice. Sit back and let them go through a few. Watch for problems and interrupt if needed. Break on sixth Every 35 maps during the real test, this break message will come up. This gives you a chance to rest your eyes, shift your position, turn your neck, and such. I'll also ask you a question for our background information. I'll ask you the first background question now: (Q 1 on form) When you are ready to continue with the test, click anywhere on that break message. Right now we are continuing with practice maps. Continue to end of practice Do you have any questions? Click on that Start Test button to begin the real test. 132 BOTH Exit: When they get to 'End' card, rejoice Open curtains EXPZ Tell me how you went about making your decisions. What strategies did you use? BOTH tell them a little about our objectives With this experiment, we're studying the perceptual effect called simultaneous contrast. Surrounding colors affect the colors you perceive. Contrast usually increases the difference between colors. If you have a gray, for example, surrounded by light colors, the gray will look darker. We're trying to predict how much color change can occur in the computer displays so we avoid designing confusing computer maps. Do you have any comments or questions about the test? Record on form or provide more explanation One last formality associated with paying you; I need you to sign this form showing you received the 10 dollars. Give them $10 and have them sign voucher Thanks Bye Back-up data file to disk Put data file into "Subject Data " folder on hard disk Fill out log sheet. Flip sign Prepare for next subject 133 The consent form signed by all subjects: CONSENT 1. I have freely consented to take part in a scientific study being conducted by Cynthia Brewer and her assistants. I will be participating in a study on the design of maps for computer displays. 2. The study has been explained to me, and I understand the explanation that has been given and what my participation will involve. 3. I understand that I am free to discontinue my participation at any time without penalty. I understand that the expected length of my participation is approximately one hour. 4. I understand that the results of my participation in the study will be kept in strict confidence, as will those of all other individuals participating. In other words, all participants will remain anonymous in the reporting of results. Within these restrictions, results of the study will be made available ' at my request. 5. I understand that my participation in the study does not guarantee any beneficial results to me. My participation will have no influence on my grades in coursework or my rating as an employee. 6. I understand that, at my request, I can receive additional explanation of the study after my participation is completed. Signed: Date: This page and the next show the background information form used for both experiments. The dashed breaks on the second page group the questions asked at each of the four test breaks during an individual session. The primary purpose of these breaks was to distract the subject from their somewhat monotonous task and to provide a conversational interlude with the test administrator. I wanted to inspire care in performance of the tasks, but avoid having the subject feel uncomfortable by watching over them without ever saying a word. The response data were astoundingly clean so this strategy may have worked. The data gathered with the background questions is not used in this research, other than to provide a general summary of subject gender, age and student or non-student status. Subject Number: Background Information Age: Gender (M/F): Will you be wearing corrective lenses while viewing the computer screen? Circle one: glasses contacts neither If you wear tinted lenses, please describe the tint: Color Vision Test Plate Response 1 135 Subject Number: Background Information 1. Have you studied color or color perception? (If yes:) Please describe your experience: 2a. In what ways do you commonly use computers? 2b. What types of color computer graphics are you familiar with? 3a. What types of maps do you commonly use? 3b. What types of computer-displayed maps are you familiar with? 4a. What is your occupation (major, if student)? 4b. How did you find out about our experiment? 4c. Have you participated in other research experiments? (If yes.) Briefly describe the experiment(s): APPENDIX B APPENDIX B Experiment 1 Color Information The extended table in this appendix provides a listing of the RG YB J, L x y, and R G B specifications for all Experiment 1 test colors. 136 137 Table 3.1 Experiment 1 Color Specifications Distance Test Medium-lightness opponent hues RGB in 100's Label Hunt: cd/m2 CIE: (max = 65535): RG YB J L x y R G B 0.0 N -0.01 0.01 72.15 32.2 .278 .302 445 420 415 2.5 R 2.50 0.00 72.21 29.7 .341 .270 655 315 410 20 1.99 0.00 72.21 30.2 .329 .276 625 345 420 1.5 1.50 0.00 72.13 30.6 .317 .282 595 370 420 1.0 r 0.99 -0.01 72.11 31.1 .304 .288 545 390 420 0.5 0.51 0.01 72.20 31.7 .292 .295 510 415 430 -2.5 G -2.51 -0.01 72.15 32.4 .217 .331 150 495 410 -2.0 -2.01 0.00 72.19 32.4 .230 .325 235 485 415 -1.5 -1.49 0.00 72.14 32.3 .243 .319 305 480 420 -1.0 g -0.99 0.00 72.17 32.3 .255 .313 360 465 420 -0.5 -0.50 0.02 72.20 32.3 .267 .308 415 450 425 3.0 Y -0.01 3.00 72.20 30.9 .411 .483 510 410 065 2.5 0.01 2.50 72.13 31.0 .387 .454 510 425 150 2.0 -0.01 2.00 72.13 31.2 .363 .425 495 425 200 1.5 -0.01 1.50 72.12 31.4 .341 .395 480 415 245 1.0 y -0.01 1.00 72.18 31.7 .319 .364 475 430 305 0.5 -0.01 0.50 72.13 31.9 .298 .333 465 430 360 -3.0 B 0.02 -3.02 72.17 30.9 .214 .188 390 360 655 -2.5 0.01 -2.51 72.12 31.1 .223 .206 400 385 620 -2.0 0.00 -1.99 72.17 31.4 .233 .225 415 400 580 -1.5 0.00 -1.51 72.16 31.6 .243 .243 410 395 520 -1.0 b 0.00 -1.01 72.15 31.8 .254 .262 420 405 485 -0.5 0.00 -0.49 72.14 32.0 .266 .282 440 415 450 Direction Test Surrounds R, G, Y, B above and secondaries: YR 2.12 2.12 72.15 29.7 .426 .376 655 340 205 RB 2.12 -2.11 72.19 29.7 .266 .196 560 285 580 BC -2.12 -2.10 72.14 32.0 .194 .240 150 460 570 GY -2.13 2.12 72.17 32.0 .282 .452 260 480 220 Table 8.1 (continued) Direction Test Comparisons RG YB I 138 L X Y RGB in 100's R G Comparison choices in color directions from 1' center (e.g. r-yr is 1.0 unit in the yellow-red direction of 45° from the red center r) r-y 1.00 0.50 72.13 31.0 .325 .313 r-yr 1.36 0.35 7215 30.7 .328 .300 r-r 1.50 0.00 72.13 30.6 .317 .282 r-rb 1.36 -0.34 72.17 30.8 .301 .269 r-b 1.00 -0.50 72.11 31.1 .287 .267 rbbg 0.66 -0.34 72.15 31.5 .284 .278 r-g 0.51 0.01 72.20 31.7 .292 .295 r-gy 0.65 0.36 72.14 31.4 .310 .311 Comparison choices in color directions from g center g-y -1.01 0.49 72.22 32.3 .266 .338 g-yr -0.66 0.34 72.23 32.3 .271 .326 gr -0.50 0.02 72.20 32.3 .267 .308 g-r'b -0.64 -0.35 72.14 32.2 .256 .294 g-b -1.00 -0.49 72.21 32.3 .245 .292 g-bg -1.36 -0.36 72.16 32.3 .239 .301 g-g -1.49 0.00 72.14 32.3 .243 .319 g-gy -1.35 0.35 72.17 32.3 .254 .334 Comparison choices in color directions from y center y-y -0.01 1.50 72.12 31.4 .341 .395 y-yr 0.35 1.35 72.19 31.3 .345 .376 y-r 0.51 1.00 72.17 31.3 .334 .351 y-rb 0.35 0.64 72.18 31.6 .314 .333 y-b -0.01 0.50 72.13 31.9 .298 .333 y-bg -0.36 0.66 72.21 32.1 .290 .343 yg -0.50 1.00 72.12 31.9 .298 .365 y-gy -0.36 1.35 72.14 31.7 .318 .387 Comparison choices in color directions from b center by 0.02 -0.47 69.12 28.7 .266 .282 b-yr 0.35 -0.65 69.84 29.2 .268 .270 b—r 0.51 -0.99 69.86 29.0 .263 .255 b-rb 0.36 -1.34 69.88 29.0 .252 .244 b-b 0.01 -1.50 70.79 30.1 .243 .243 b-bg -0.35 -1.36 69.86 29.3 .240 .253 bg -0.49 -0.99 69.89 29.5 .246 .269 b-gy -0.34 -0.65 69.83 29.5 .256 .280 570 590 595 570 530 510 510 530 367 400 415 405 360 325 305 335 525 525 513 430 453 455 470 410 395 405 400 380 370 380 390 410 415 405 467 460 450 452 463 475 480 482 415 427 415 380 390 420 455 470 460 430 390 380 395 425 455 465 455 420 400 245 287 318 370 372 332 450 528 520 520 Table 8.1 (continued) EfisumnceTknn 1R1: YT! J Light and dark saturated hues Dark reds 6.22 6.21 0.00 57.56 5.22 5.20 0.00 57.53 4.22 4.22 0.01 57.49 3.22 rd 3.20 0.01 57.54 2.22 2.22 0.01 57.57 fL22 1L22 (L01 E§l56 (L22 (L22 4101 lilSO Iightgpeens -2.85 -2.85 0.00 83.92 -2.35 -2.36 0.00 83.90 -1.85 -1.86 0.01 83.95 -1.35 gl -1.35 0.01 83.93 4185 ‘4L84 (100 835%) -0.35 ~0.36 -0.01 83.87 Light yellows 3.14 0.01 3.14 83.51 2264 (L00 :L63 (KL52 2.14 -0.01 2.14 83.52 1.64 yl 0.01 1.65 83.51 1.14 0.01 1.14 83.47 (L64 (101 (L63 83:94 0.14 -0.01 0.14 83.51 Dark blues -7.26 0.10 -7.26 44.52 4526 -0(LZ (4&26 (QLSZ -5.26 0.03 -5.26 44.62 ~426lxl (L00 4L28 (9&52 -3.26 0.00 -3.25 44.65 -2.26 -0.02 -2.24 44.61 -1.26 -0.01 -1.26 44.60 139 15.3 15.7 16.1 16.6 17.1 17.6 18.1 48.0 47.9 47.9 47.8 47.7 47.6 44.8 45.1 45.4 45.7 46.0 46.5 46.9 8.9 9.1 9.3 9.4 9.6 9.7 9.8 .415 .395 .375 .353 .278 .210 .249 .261 .417 .394 .372 .351 .329 .307 .161 .166 .175 .185 .199 .215 .248 .259 .270 .282 .294 .339 .332 .326 .320 .313 .307 .458 .432 .374 .343 .312 .076 .118 .143 .173 .206 .241 RGB in 100's R G 655 060 620 140 566 185 523 225 475 260 420 293 350 325 0 635 235 605 330 600 400 575 470 564 530 555 655 515 655 540 620 530 620 545 605 540 600 555 585 550 190 070 180 145 180 175 190 198 200 210 210 225 220 230 320 325 320 325 325 330 510 513 505 515 520 175 235 510 655 590 510 455 395 345 295 Table 8.1 (continued) 140 Distance Test RGB in 100's RG YB J L x y R G B lightness 90.7 W 0.00 0.00 90.71 58.5 .282 .304 655 615 595 86.1 0.01 -0.01 86.07 50.9 .281 .303 605 575 560 80.3 0.00 0.00 80.27 424 .280 .303 530 500 485 74.5 1 0.01 0.00 74.54 35.0 .279 .302 485 455 445 68.7 0.00 0.00 68.68 28.4 .277 .301 430 410 403 62.9 -0.01 0.01 62.94 22.8 .275 .300 380 365 360 57.1 m 0.01 0.01 57.10 17.9 .273 .298 325 310 310 51.3 0.01 0.00 51.30 13.8 .269 .295 280 270 275 45.5 0.00 0.00 45.51 10.4 .264 .292 239 240 253 39.7 d (0.00 0.01) 39.66 7.6 (.258 .288)‘ 195 195 205 33.9 (0.00 0.00) 33.90 5.4 (.250 .282)‘ 160 160 168 28.2 (0.01 0.01) 28.17 3.7 (.239 .274)* 131 131 138 22.4 (0.01 0.01) 22.36 2.4 (.223 .262)‘ 100 100 105 16.9 K (-0.03 0.00) 16.87 1.5 (.199 .245)‘ 060 060 063 Grays in control-surround patchwork Grays have a slight blue-green color. They were grays available in the system ‘ color table and were used to avoid a distracting flash of colors when color tables between displays were changed. L,x,y is mean of 20 measurements of gray within patchwork. Grays 1, 2, 4, 6, 8, 10 are immediately adjacent to comparison hexagons and center test hexagon with control surround. RG YB J L x y R=G=B 1 on -0.19 92.93 62.4 .273 .299 65535 2 -0.24 -0.19 89.04 55.7 .272 .299 61166 3 -0.19 -0.16 84.72 48.8 .273 .299 56797 4 -0.20 -0.10 75.30 35.9 .272 ~ .300 48059 5 -0.21 -0.08 70.61 30.5 .271 .300 43690 6 -0.14 0.02 60.46 20.6 .271 .301 34952 7 (0.02 0.19) 42.08 8.7 (.269 .302): 21845 8 (0.09 0.30) 35.54 5.9 (.268 .303): 17476 9 (0.50 0.63) 19.35 1.8 (.268 .305): 08738 10 (0.83 0.72) 12.82 1.0 (.275 .305): 0 " Colors too dark for accurate measurement of x,y chromaticity with Minolta Chroma Meter (the low luminance measures (L), however, are within meter’ 5 range) APPENDIX C APPENDIX C Experiment 1 Frequency Response Data The tables of this appendix summarize frequencies of response for all test displays from Experiment 1. Each row provides numbers of responses for a single comparison color. For each display, six comparison colors were offered, and choices that were not selected by any of the subjects are marked with a dash H. In a few cases choices were duplicated and these are marked with a repeated dash (-). The columns list responses for individual test displays (N=30 in each column). The first row of column labels lists the center and inducing surround colors. For example, I W is a light center with a white surround and r G is a red center with saturated green surround (see Appendix B). The second row of column labels describes the size of the hexagons in the display grid: 11 for 11 mm, 6 for 6.5 mm, and 4 for 4.5 mm. The letter portion of these labels designates the type of test: d for the primary distance tests, c for control, w for white outlines, k for black outlines, and v for heterogeneous inducing surrounds (see Chapter 2 methods description for explanation of types and supporting figures). 141 142 Table C.1 Frequency Response Data for Lightness Distance Test Lightness Distance Test: Light Center lcontrol l W lm —G1_(fi._911_d§__fi__h§__b 96 d4 d6 90.7w - 86.1 - - - - - - - 80.3 5 8 5 1 - 1 1 3 7 12 74.5 1 21 22 25 8 6 7 18 21 16 68.7 4 - - 14 16 15 20 8 2 2 62.9 - - - 6 7 7 2 1 - - 57.1 m - - - 1 - - - - - 51.3 - 1 (cont'd) Id 1 K d4 dll d4 ® d1] v4 v1] 90.7 W l 1 1 1 - 86.1 1 - 5 6 S 3 - 80.3 8 14 17 16 19 16 28 74.5 I 19 14 4 7 5 8 2 68.7 1 2 3 - - 1 - 62.9 1- - - - - 1 - 57.1 m - 51.3 143 Table C.1 (continued) Lightness Distance Test: Medium Center 80.3 -- - 74.5 I - - - 68.7 - - - 62.9 1 1 3 57.1 m 25 27 26 16 17 17 9 Z) 51.3 2 2 3 7 11 11 17 6 45.5 2 - 1 39.7 d - - 33.9 - - 1.98“! H o 'er5 0| pa 0 (cont'd) mK d4 fi dll W4 k4 v6 86.1 80.3 - - - - 74.5 I 1 68.7 3 6 3 1 62.9 1 57 .1 m 7 10 1 12 51.3 1 1 - 5 - 4 45.5 - - 39.7 d - 33.9 on and U) N 0‘ H N @t-IO‘t-l \l \1 144 Table C.1 (continued) Lightness Distance Test: Dark Center d control d W d 1 CL (6 CH d4 (5 d1] w11 Jkll v11 d4 CB 62.9 - 57.1 m - - 51.3 2 - - 2 - 2 - - 1 45.5 5 3 - 3 2 3 2 2 1 3 1 39.7 d 19 25 24 10 18 13 15 Z) 22 18 18 33.9 4 2 6 12 9 14 13 6 7 8 9 28.2 - - 3 1 - - - 1 1 22.4 - - - - - - - - - 16.9 K - - (cont'd) d m d K :5 dll d4 d6 dll v4 62.9 1 - - 57.1 m - 2 1 - 51.3 - - 7 7 7 1 45.5 5 7 7 15 15 7 39.7 d 22 20 12 6 8 14 33.9 2 3 2 1 - 7 28.2 - - 1 22.4 - - 16.9 K 145 Table C.2 Frequency Response Data for Hue Direction Test 2.5 R 2.0 1.5 1.0 r 0.5 0.0 N -0.5 (cont'd) 2.5 R 2.0 1 .5 1.0 r 0.5 0.0 N -0.5 -2.5 G -2.0 -1.5 -1.0 g -0.5 0.0 N 0.5 (cont'd) -2.5 G -2.0 -1.5 -1.0 g -0.5 0.0 N 0.5 IHBgotg d1] 12 17 8101 v6 (111 Iggpr 12 18 dll W646 13 17 .8“. Inmur dll 146 Table C.2 (continued) y control y Y yb —_d__fi_fll_du_dll__fl_fl_fll_d4__d§_dll_ 3.0 Y - - - - - 2.5 - - - - - - - - - 2.0 - - - - - — - - - 2 1 1.5 6 - - - - - - - 3 8 11 5 1.0 y 24 29 25 - - r 2 3 8 19 17 0.5 - 1 5 20 20 22 20 21 19 1 l 0.0 N 10 10 7 8 6 - -0.5 - - (cont'd) y B d4 (5 dll v4 3.0 Y — - 2.5 - - - 2.0 1 - - 1 1.5 6 15 8 11 1.0 y 22 14 22 17 0.5 1 1 - 1 0.0 N - - -0.5 - b control b B b y c4 (15 cll d4 d6 dll W6 11 -3.0 B - - - - - - -2.5 - - - - -- - - - - - - -2.0 - - - - - - - 3 1 1 -1 .5 l 1 - - - - - - 9 l7 6 -1.0 b 28 28 30 1 - - - 1 17 12 23 -0.5 1 1 - 7 8 8 1 - - 0.0N 22 22 28 23 21 147 Table C.3 Frequency Response Data for Light and Dark Hue Distance Test rd control rd R rd G c}: c6 (:11 (14 i (111 wll l_<11 v4 d4 (5 4111 6.22 - - 5.22 - - - - - - - - - - - - 4.22 2 - - - - - - - - 1 3 - 322 rd 27 30 30 7 4 6 24 9 25 29 24 23 2.22 1 - - 22 25 24 5 21 5 - 3 7 1 .22 - - - 1 - 1 - 1 - - - - - 0.22 - - - - - - - - - g1 control gt G g1 R 04 (6 d1 d4 CB dll d4 CE dll W4 k4 v11 -2.85 - - - - - - - - - - - - ~2.35 - - - - - - - - - - - ~ -1 .85 1 - - 1 - - - 1 2 - 1 -1.35 gl 27 28 28 7 6 4 20 23 22 23 22 12 -0.85 - 1 2 23 22 26 10 7 7 5 8 17 -0.35 - - - - l - - - - - - - yl control yl Y yl B 04 (6 $11 d4 d6 v4 d4 d6 dll w11 kl l 3.14 - - - - - 2.64 - - - - - - - - - - - 2.14 2 - - - 1 4 4 3 - 1.64 yl 24 27 26 3 2 18- 13 22 19 25 27 1 .14 5 1 4 21 21 12 15 4 7- 2 3 0.64 - - - 6 7 - 1 - - - 0.14 - - - - - - bd control bd B bd Y 04 (6 £11 d4 (E d1] d4 (3 dll W6 16 v1 1 -7.26 - - - - - - - -6.26 2 - - - - - - 1 - 1 - - -5.26 5 1 3 1 1 2 - 6 5 1 4 4.26 bd 20 28 18 12 7 14 15 15 22 19 19 25 -3.26 3 1 5 12 20 14 12 13 2 5 10 1 -2.26 - - - 3 2 l 1 1 - - - - -1 .26 148 Table C.4 Frequency Response Data for Hue Direction Test Center: r Surround: 18 BC G GY Y YR R RB r-y 17 4 2 - - r-yr 2 23 4 4 - - - r-r 1 - 17 3 - - r-rb - 1 7 16 1 - r—b - 5 7 1 - - r-bg - 2 18 18 4 r-g - 1 - 4 10 26 1 r-gy 10 1 1 - 27 Center: g Surround: B BG G GY Y YR 1R RB g-y 14 - - 8 7 g-yr 15 10 - 4 1 5 4 g-r - 16 25 8 - 2 - g-rb - 4 5 16 2 3 g-b - 2 22 16 1 g-bg - - 5 5 3 - 8'8 ' ' ‘ ’ 4 9 g-gy 1 - - 4 14 Center: y Surround: B B6 G GY Y YR R RB y-y 10 9 2 1 m y-yr 15 16 19 - data y-r 1 3 9 5 - - y-rb 4 l - 23 - - - y-b 1 - 2 22 4 - y-bg - - 8 15 - 7'8 ‘ ' ’ 9 9 my - - - 2 20 Center: b Surround: B BG G GY Y YR R RB b-y 26 9 - - 1 8 b-yr 2 19 13 7 1 2 - b-r - 1 14 10 - - b-rb 1 3 13 1 1 b-b - - - 7 1 - b-bg - - 11 20 5 1 b-g - - 10 4 14 2 b—gy - - 4 8 19 APPENDIX D APPENDIX D Program Code for Hunt Model of Color Appearance The Hunt model was programmed in HyperCard and was used interactively. Program code was associated with buttons (graphics that are clicked on with the mouse to run their code or script) and with cards (the individual screens displayed as the program is used). Descriptions were also listed in fields (text blocks that are scrolled to accommodate lengthy discussions). In this appendix, each card is shown and the text and scripts associated with that card are listed in the pages that follow the card. Altogether there are six cards: 1. Hunt’ 5 Model of Color Appearance 2. Input Template 3 Calculate RGB Cone Absorption Values 4. Calculate Cone and Rod Responses 5 Intermediate Measures of Appearance 6 Sample Appearance Measures 149 150 2...; 8:222 :5 29:8 2.. £3 .2 m2:8oE 2:8 :o 282 2 :8: us. .222. a 2 .82 2: .2: 82m: 22. .2 n8: 2: L. :. Bo .2: 82:2 H. 2.8 222.8 5.3 .232. 2: 2 882. 8 _...s 2...: 8:222 cc: 2:28 2: £3 .2 822, 95:88:00 :2...» .b. :_ 9.0:: «2:8 22. 228.332 .222. a 2 .82 8.92.; b 2...: 8:222 2.. .2 8.382: 2: 5?. :_ o:.u:o 8:3: 222.; ”2, ”2.2 8.2.8.: >_.8..mE2m>m a :_ .o .:52... now: 2: 82:: 292.2. 8:3: 5.28:0 .8558 .o .85 :. 22:23.3: 82:8... 2 E202: 2:. :. :8: 2: .2: 8.92.2, 2.. .0 22.2.3 2..» can; 23:. 392w "832.; 3 32.555 88:222.. .>.:o 20.8 :222 .2 2m 82.2860 camp .2:.>> .3265 255.0 .3 32228.2 8:282? .200 .o .322 «.52.. DJ 1'" 151 Introductory Notes Field Programmed by Cynthia Brewer, Winter 1990 Calculations are for related colors only (not isolated color patches). References: Hunt, R.W.G., 1987, "Measuring Colour," Chichester: Ellis Horwood, Chapter 8 (pages 146 to 173). Calculations are numbered with the step numbers from the above description of the model. The steps have been reordered to allow a logical program flow, but they retain their original step numbers for ease of reference to the original formulation. Page numbers in the documentation of this program are for the above primary reference unless otherwise noted. Page references followed by "CRA'87" are to the following article: Hunt, R.W.G., 1987, A Model of Colour Vision for Predicting Colour Appearance for Various Viewing Conditions. "Color Research and Application" 12(6): 297-314. Page references followed by "CRA'82" are to the following article: Hunt, R.W.G., 1982, A Model of Colour Vision for Predicting Colour Appearance. "Color Research and Application" 7(2): 95-112. ’New Input Stack’ Button on mouseUp lock screen go to card "input template" doMenu Copy Card doMenu New Stack... put the short name of this stack into newStack if newStack <> "hunt code" then doMenu Paste Card go to first card doMenu Cut Card unlock screen tabKey else - 'New Stack' dialogue box was cancelled go to first card unlock screen end if end mouseUp 152 Summary of Variables Field This summary of the variables that are used in this program is organized approximately in order of appearance. Generally, Hunt's variable names are used directly or in a systematically modified form: W: Variable names ending in "W" are measures for the reference white. b: Variables local to a handler (subprogram) have names ending in "b" when corresponding values for both the sample and reference white will be passed to the handler with separate calls . f: Names that end in "f" are used for field names that are local to a handler and used to record on cards measures for both the sample and reference white. Chromaticity coordinates for sample, reference white, and local use in handler x y, xW yW, xb yb Luminance factors: sample, ref W, local bigY, bigYW, bing Photopic luminance in cd/mAZ: sample, ref W, local L, LW, Lb Luminance in cd/m"2 of adapting background: photopic, scotopic LA, LAS Factor for calculation of LAS scotop Chromatic induction factor N c Brightness induction factor Nb Name entered on current input card label Full card name for current input card dataCard XYZ tristimulus values: local bigXb. bing. biga 153 Absorptions of three cone types: sample, ref W, equal-energy illuminant (SE), local, field R G B, RW GW BW, RE GE BE, Rb Gb Bb, Rf Gf Bf Stimulus response functions: cones, rods Fpr, FNfs Stimulus response function local arguments: cones, rods pr,fsC Luminance-level adaptation factors: photopic, scotopic FL, FLS Measures of purity of the adapting illuminant for each cone type hr hg hb Retinal-stage chromatic adaptation factors for each cone type FR FG FB Cortical-stage chromatic adaptation factors for each cone type RD GD BD Responses after adaptation for each cone type: sample, ref W, local, field Ra Ga Ba, RaW GaW BaW, Rab Gab Bab, Raf Gaf Baf Color difference signals: sample, ref W, local, field C1 C2 C3, C1W C2W C3W, C1b C2b C3b, le C2f C3f Rod response after adaptation Sa Photopic achromatic response Aa Total achromatic response: sample, ref W, local, field A, AW, Ab, Af Blue-Yellow response t Red-Green response P Hue angle (360 degrees): sample and local, ref W, field hs, hsW, hsf 154 Hue angles for unique hues (constants) Rhs Yhs Ghs Bhs Eccentricity factors for unique hues (constants) Res Yes Ges Bes Constants for unique hue with nearest lower hue angle: hue angle, eccentricity factor hl, e1 Constants for unique hue with nearest higher hue angle: hue angle, eccentricity factor h2, e2 Eccentricity factor es Low-luminance tritanopia factor Ft Blueness-yellowness: sample, ref W, local, field bigMBY, bigMBYW, bigMBYb, bigMBYf Redness-greenness: sample, ref W, local, field bigMRG, bigMRGW, bigMRGb, bigMRGf Colorfulness: sample, ref W, local, field M, MW, Mb, Mf Relative blueness-yellowness with more convenient numbers: sample, refW mBYc, mBYcW Relative redness-greenness with more convenient numbers: sample, refW mRGc, mRGcW Brightness: sample, ref W Q, QW Lightness of sample I Chroma of sample C 155 Intermediate variables used to split calculations (defined immediately before use) divsum, conesum Variables used to scale plot of mRG, mBY, and Q for sample: RGpt Bth th Scale circle proportional to Q for plotting radius Position circle on plot: bottom-left, top-right (RG, BY coordinates) RGbl BYbl, RGtr BYtr Flag to limit measures calculated and output for reference white (contains "yes" or 00:10") refW 156 0 0 a 32. Q5225 E: 22; 2.2.3 05 55828: 9.5.84 @3258. l :28» H.32.. 8.22:8 80:22.3 3030.5 _ m. 2, 5 650.932... 2:83. E 7 a: z... 2...; 8:222. ".9528. T :02 ._ ”2.50m «852:3 , 2020:: 03780.. 08L, > > x ”22; 8:222 a 03.00— 0n~._ 35.— > >x .2028 "0.0.0:“. 00:052.... 0:: «3050.000 3.0302020 D.2 Card 2 F' 157 ’open file’ Button on mouseUp global outfile put background field "Outfile name" into outfile open file outfile write outfile 6:6: the date to file outfile write return to file outfile write "xW yW LW LA scotop" to file outfile write return to file outfile write background field "xW" & " " to file outfile write background field "y " 6t " " to file outfile write background field "LW" 8: " " to file outfile write background field ""LA 8: " " to file outfile write background field "scotop" to file outfile write return to file outfile write "name, RG, YB, I, L, x, y" to file outfile write return & return to file outfile end mouseUp ’Run’ Button on mouseUp global x, y, bigY, xW, yW, bigYW, L, LW, LA, LAS global label, dataCard, dataStack put field "L" into L put field "LW" into LW put (L / LW " 100) into bigY put bigY into field "bigY" put field "x" into x put field "y" into y put field "xW" into xW put field "yW" into yW put field "bigYW" into bigYW put field "LA" into LA put field "LAS" into LAS put field "Name of Sample" into label set name of this card to "input," && label put the long name of this card into dataCard put the long name of this stack into dataStack set hilite of. background button "Run" to true go to card "RGB calc" of stack "hunt code" CALLSl end mouseUp IL. 158 ’Calc LAS’ Button on mouseUp -- Disagreement exists in sources on the 2.26 multiplier used to - calculate LAS. 2.46 appears in "Measuring Colour" and 2.26 appears - in the Hunt 1987 CRA reference. Use of 2.26 is consistent with - use of the 2.26 divisor in the FLS equation (Step 15). - Use of 2.26 in this calculation produces correct output for example - calculations in "Measuring Colour," although neither 2.26 or 2.46 multipliers —- produce results matching the LAS values reported for those example colors. put field "LA" into LA put field "scotop" into scotop put (LA " scotop " 2.26) into field "LAS" set hilite of background button 2 to true end mouseUp 159 .8355: 5.3.9628 ~88on 32630 03625300 93255 H 0...: «:00 g “has m5 05 to Emma 839. mwm «836.5 m «mo. 3 mm 388.69 3m m02a> 93.3.8 883.? 0 a 638mm «2.8 833 m0 mm ”Emcee-g: 3.55-335 2.83.8. 333.3 30 3: “BE; 8:228”. cozeomn< 230 m0: 32330 D.3 F' 160 RGB Calculations (XYZRGB) This portion of the program is a ’card script’ that is called by the ’Run’ button. on CALLSI global xW, yW, bigYW, x, y, bigY - Calc for reference white XYZRGB xW, yW, bigYW, "RW", "GW", "BW" - Calc for equal-energy illuminant (SE) XYZRGB 1 / 3, 1 / 3, 100, "RE", "GE", "BE" - Calc for sample XYZRGB x, y, bigY, "R", "G", "B" send mouseUp to card button "continue" end CALLS] CALCULATE CONE ABSORPTION VALUES on XYZRGB xb, yb, bing, Rf, Gf, Bf — 1. Calc CIE 1931 XYZ tristimulus values (p. 168): bigXb bing biga put ((xb " bing) / yb) into bigXb put ((1 - xb - yb) " bing / yb) into biga - 2. Calc cone absorptions: Rb Gb Bb - RGB represent amounts of radiation usefully absorbed per unit area of the -- retina by the three cone types in a given state of adaptation. - The calculations are based on the spectral sensitivity functions of - the CIE 1931 Standard Colorimetric Observer. (pp. 147, 148, 168) put ((0.40024 " bigXb) + (0.7076 * bing) - (0.08081 " biga)) into card field Rf put ((-0.2263 * bigXb) + (1.16532 * bing) + (0.0457 * biga)) into card field Gf put (0.91822 " biga) into card field Bf end XYZRGB 33mm... 63»; 280 Buo 5392 3am 2.33.. 533.. no No 161 333.6 cm d osczceup ”SE; .3. a2 £29m 62.92% 5.8 new .956 oszoEow 0329?. 5:333 .23 SE; 8:222 2 39532 280 ”anmm .2 229m 855:6 8.8 05.... .chm oszoEom 032on ”conning 8cm macaw 2 88039 6:60 momconmom no: can 230 923.60 DA F' 162 Cone and Rod Response Calculations (ADAPT, CON EROD) This portion of the program is a ’card script’ that is called by the preceding ’Continue’ button on CALLSZ global R, G, B, RW, GW, BW, bigY, bigYW ADAPT CONEROD R, G, B, bigY, "Ra", "Ga", "Ba", "A", "C1", "C2", "C3" CONEROD RW, GW, BW, bigYW, "RaW", "GaW", "BaW", "AW", "C1W", "C2W", "C3W" send mouseUp to card button "continue" end CALLSZ - STIMULUS RESPONSE FUNCTIONS: - Physiologic evidence supports the use of hyperbolic functions for - these two calculations. (pp. 148, 168 and pp. 297, 298 CRA'87) - CON ES function Fpr pr return (40 " ((pr"0.73) / (pr"0.73 + 2)) + 1) end Fpr -- RODS - Lower constants and exponents than in cone function reflect the lower - contrast and brightness mediated by rod vision. (pp. 160, 170) function FNfs fsC return (30.5 ” (30 / (fsC + 30)) * (fsCAO.56 / (fsC"056 + 0.16)) + 0.61) end FNfs - CALCULATE ADAPTATION FACTORS USED IN BOTH REFERENCE WHITE AND - SAMPLE CALCULATIONS on ADAPT global LA, LAS, RW, GW, BW, RE, GE, BE, FL, FLS, FR, FG, FB, RD, GD, BD .. 4. Calc luminanceblevel adaptation at a retinal processing stage: FL -- (SLA is approximation of luminance of reference white) (pp. 149, 150, 168) - The cube-root relationship used for FL is common in research relating - visual response and intensity (p. 288 CRA'87). put «(100 * (5 .. LA» / ((5 * LA) + 1005))«1 / 3) + 0.001) into FL 163 - 5. Calc measures of purity of the adapting illuminant". hr hg hb - Calc chromatic adaptation factors for retinal-stage processing: FR ES FB —- FR, PG, and FB represent the effects that adaptation to colors of - lights becomes less complete as illuminant purity increases and - more complete as luminance increases. (pp. 149, 152, 168) - RE, GE, BE are the RGB values for the equal-energy stimulus (SE). - SE is used as an anchor in the calculation of FR, FG, FB because - subjective neutral points tend to be displaced from the illuminant - point toward SE when plotted in the chromaticity diagram. (p 302 CRA'87) put ((RW / RE) + (GW / GB) + (BW / BE)) into divsum put ((3 " RW / RE) / divsum) into hr put ((3 " CW / GE) / divsum) into hg put ((3 " BW / BE) / divsum) into hb put (1 + LA"(1 / 3)) into LA3rd put ((LA3rd + hr) / (LA3rd + (1 / hr))) into FR put ((LA3rd + hg) / (LA3rd + (1 / hg))) into FG put ((LA3rd + hb) / (LABrd + (1 / hb)» into FB —- 6. Factors for calculation of extremely rapid chromatic adaptation occurring -- at a later cortical processsing stage: RD GD BD - These factors account for the Helson-Iudd effect and the effect of - discounting the color of the illuminant (color constancy). - (pp. 149,153, 168) - The Fpr and FL functions establish approximately square-root and - cube-root relationships respectively. Their combination relates - colorfulness (M) to the sixth-root of illumination level, which is -- consistent with experimental evidence. (p. 151) put ((Fpr(0.2 * FL * FG)) - (Fpr(0.2 * FL * FR)» into RD put 0.0 into GD put ((Fpr(0.2 " FL " PG» - (Fpr(0.2 " FL * EB») into BD - 15. Calc scotopic luminancelevel adaptation factor: FLS - (pp. 161, 170) - LAS / 2.26 is used instead of LAS in the FLS equation because - LAS / 2.26 equals LA for SE (p. 306 CRA'87). put ((100 ‘ (5 " LAS / 2.26) / (5 " LAS / 2.26 + 1005))“(1 / 3) + 0.001) into FLS end ADAPT 164 -- CALCULATE CONE AND ROD RESPONSES FOR THE SAMPLE AND REFERENCE WHITE on CONEROD Rb, Gb, Bb, bing, Raf, Gaf, Baf, Af, le, C2f, C3f global bigYW, FL, FR, FG, FB, RW, cw, BW, RD, CD, so, FLS -— 3 6t 7. Calc cone responses after adaptation: Rab Gab Bab -— (pp. 149, 168) put ((Fpr(FL * FR " (Rb / RW))) + RD) into Rab put ((Fpr(FL " FG " (Gb / CW») + CD) into Gab put ((Fpr(FL " FB " (Bb / BW))) + BD) into Bab put Rab into card field Raf put Gab into card field Gaf put Bab into card field Baf - 16. Calc rod response after adaptation: Sa - bing / bigYW is the photopic equivalent used to approximate scotopic - luminance relative to the reference white scotopic luminance. - (pp. 161, 167, 170) put (FNfs(FLS " (bing / bigYW))) into Sa :- 8. Calc color difference signals: C1 C2 C3 - Calc photopic achromatic signal: Aa - Notes: C1=C2=C3=0 is the criterion for achromatic perception. - C1:C2:C3 in constant ratio yields constant hue perception. - Factors 2 and 20 in the calculation of Aa are based on the relative - abundance of the three cone types (R:G:B is 40:20:1). -— (pp. 154, 169) - 17a. Calc total achromatic response: Ab (pp. 161, 170) put (Rab - Gab) into card field C1f put (Gab - Bab) into card field CZf put (Bab - Rab) into card field C3f put ((2 " Rab) + Gab + (Bab / 20)) into Aa put (Aa + Sa) into card field Af end CON EROD 5L 165 fl 025:3 w {32.0 338.- 628.- 25 30m: Em: ”$6523.00 new .0-m .>-m 0:322 32.. No.9? on... 0:55 oococouom hmhoovd nanomvd hmmooof .2 0m: >0: 6852860 new .Oi .>-m 388.98 a... 6.92 on: oEEam 60:263.? .6 «2:235. 26.62.52... D.5 F' 166 Intermediate Measures of Appearance (INTAPP) This portion of the program is a ’card script’ that is called by the preceding ’Continue’ button on CALLS3 global B, BW, bigY, bigYW, L, LW, C1, C2, C3, C1W, C2W, C3W INTAPP B, bigY, L, C1, C2, C3, "hs", "MBY", "MRG", "M" INTAPP BW, bigYW, LW, C1W, C2W, C3W, "hsW", "MBYW", "MRGW", N M w u send mouseUp to card button "continue" end CALLS3 - CALCULATE INTERMEDIATE CORRELATES OF APPEARANCE on INT APP Bb, bing, Lb, Clb, C2b, C3b, hsf, MBYf, MRGf, Mf - The following calcs produce unique hues under these conditions: - Red C1=C2, Yellow C1=C2/11, Green C1=C3, Blue C1=C2/ 4 (p. 154) -- 9. Calc hue angle (a correlate of perceived hue): hs (pp. 155, 169) - The R-G axis, p (for protan), is the average of the color's - difference from unique red (C1 - C2) and the difference from unique - green (C3 - C1). The average simplifies to 0.5(C3—C2). -- The B-Y axis, t (for tritan), is calculated using only the color's - difference from unique yellow (C1 - C2 / 11) because it is more -- sharply defined than unique blue. The '4.5' divisor is - approximately equal to SQRT(20) and is used to account for the - lower population of B cones. - (pp. 101, 102 CRA'82) put ((-0.5) * (C3b - C2b) / 4.5) into t put (C1b - C2b / 11) into p ifp=0andt>=0then put 90 into hs elseifp=0andt<0then put 270 into hs else put ((Atan(t / p)) * (180 / pi)) into hs - Arctan converts from cartesian BY,RG coordinates to the angular - component of a polar coordinate expression (hs,M) end if 167 ifp<0then put(hs+180)intohs elseift<0andp>0then put(hs+360)intohs endif put hs into card field hsf - Set constants for calculation of es (calculation of H at Step 10 - has been omitted because H is not used in the induction model). put 20.14 into Rhs put 0.8 into Res put 90 into Yhs put 0.7 into Yes put 164.25 into Ghs put 1.0 into Ges put 237.53 into Bhs put 1.2 into Bes if hs >= 0 and hs < Rhs then put (hs + 360) into hs end if if hs >= Rhs and hs < Yhs then put Rhs into h1 - hs and es for unique hue with nearest lower put Res into e1 - value of hs put Yhs into h2 - hs and es for unique hue with nearest higher put Yes into e2 - value of hs else if hs >= Yhs and hs < Ghs then put Yhs into h1 put Yes into e1 put Ghs into 112 put Ges into e2 else if hs >= Ghs and hs < Bhs then put Ghs into h1 put Ges into e1 put Bhs into h2 put Bes into e2 else ifhs>=Bhsandhs5 fi Dimm— 2. a: a >+ «Saunas. ooze-3&3 29:90.. 8.8 ":52 8.5.3.5 3.. rm; 25...: 2...- 70+: 25...: 35.52:? 3...; 00:230.”. Eds a: 3.5.3: mmdw no. 3.5.3.5 o "36.5 .n +> 252.: md "some. .o i 25...: m md Figure D.6 Card 6 171 Final Results (FINAL) This portion of the program is a ’card script’ that is called by the preceding ’Continue’ button on CALLS4 global label, dataStack, bigMBY, bigMRG, Ra, Ga, Ba, M, A, bigMBYW, bigMRGW, RaW, GaW, BaW, MW, AW - create new results card to carry data lock screen doMenu Copy Card go to last card of dataStack doMenu Paste Card set the script of this card to empty set name of this card to ("results," 6:8: label) put label into field "name of sample" unlock screen FINAL bigMBY, bigMRG, Ra, Ga, Ba, M, A, "no", "mBYc", "mRGc", "Q" FINAL bigMBYW, bigMRGW, RaW, GaW, BaW, MW, AW, "yes", "II‘BYCW", "mRGCW", "QW" end CALLS4 - CALCULATE THE REMAINING CORRELATES OF APPEARANCE FOR THE REFERENCE -- WHITE AND SAMPLE on FINAL bigMBYb, bigMRGb, Rab, Gab, Bab, Mb, Ab, refW, mBch, mRGcf, Qf global MW, AW, L - Measures that normally increase with increasing levels of - illumination: bigMBY, bigMRG, M, Mc, Q -— At extremely high or extremely low illumination, the rate of - increase gradually reduces to zero. - Measures that normally remain constant with increasing illuminantion: -— hs, H, mBYc, mRGc, s, I, C - (p. 304 CRA'87) put (Rab + Gab + Bab) into conesum put (bigMBYb / conesum * 100) into mBYc put (bigMRGb / conesum * 100) into mRGc 172 - 17b. Use total achromatic response plus colorfulness: A + M - 18. Calc correlate of brightness: Q - (pp. 162, 170) - Equal LOG (Q) differences correlate with equal steps perceived - brightness (p. 308 CRA'87) -- QW is brightness of reference white -- AW is total achromatic response to reference white - Nb is the general induction factor for brightness from a 40 degree -- surround. - Nb approximations: (p. 167) - 400 small areas in uniform surrounds - 100 normal scenes - 30 television and VDU displays in dim surrounds - 10 projected photographs in dim surrounds - 5 arrays of adjacent colors in dark surrounds - As illuminanance increases, N1 represents the increase in brightness -— and N2 represents the increase in contrast (p. 308 CRA'87). - Dr. Hunt said (personal communication April 1989) that - the Nb factor was producing inaccurate results in further model - testing. - Thus, these two equations were used when first selecting Exp I colors: - put (Ab + Mb) into Q - put (AW + MW) into QW - Initially, I wrongly omitted the N1 and N2 factors to leave all -- induction prediction to additions from my induction model, but that - omission produces a non-uniform lightness scale, which is undesirable -- (February 1991) p. 309 CRA '87 - If the values of AW and Nb are such that AW * Nb"1.23 = 21257, then - the I expression will reduce to CIE 1976 L" (with M=0 ). - Solve for Nb: put ((21257 / AW )".813) into Nb put Nb into background field "Nb" put AW into background field "AW" - Use original equations for Q and QW: put ((AW"O.5) / (5.33 * (Nb"0.13))) into N1 put ((AW"1.9) " (NbAO.362) / 2080) into N2 put (((Ab + Mb) * N1) - N2) into Q put (((AW + MW) * N1) - N2) into QW 173 - 19. Calc correlate of lightness (brightness relative to brightness - of reference white): I - I is similar to CIELUV 1976 L" when the range of LW = 100 to 10000 - cd/mAZ and Nb is 100. Maximum I is approximately 100 (pp. 162, 163, 171) - (see Q discussion for specific manipulations to set I to L‘) put (100 * Q / QW) into] - Final output set numberFormat to (##0##) put mBYc into field mBch put mRGc into field mRGcf put Q into field Qf if refW is "no" then put] into field "I" - plot rel RG, rel BY, Q put round(281 + (mRGc * 17)) into RGpt put round(185 - (mBYc * 17)) into Bth choose pencil tool click at RGpt, Bth put round(Q"0.5 " 2) into radius put (RGpt - radius) into RGbl put (RGpt + radius) into RGtr put (Bth + radius) into BYbl put (Bth - radius) into BYtr choose oval tool drag from RGbl,BYbl to RGtr,BYtr put (287 - (round(Q '* 286)» into th choose line tool drag from 483,th to 473,th choose browse tool end if end FINAL 174 ’output' Button on mouseUp global outfile, dataCard put background field "mRGc" into RG put background field "mBYc" into BY put background field "I" into] put round(I) into Ilabel go to dataCard write" " 6: ]1abel to file outfile write " " 6: background field "Name of Sample" 6: "," to file outfile write RG 6: ", " to file outfile write BY 6: ", " to file outfile write I 6: ", " to file outfile write background field "L" 6: ", " to file outfile write background field "x" 6: ", " to file outfile write background field "y" to file outfile write return to file outfile end mouseUp APPENDIX E APPENDIX E Experiment 2 Color Information The listing in this appendix summarizes the map color schemes used for Experiment 2. The map number, map title, and a summary of the center- surround pairings for each map is given. Tables list RG YB J, L x y, R G B specifications for the colors. The legend category names and approximate descriptive color names are also listed (the second set of color names for each map have rounded lightness-number prefixes). After the Experiment 2 work was completed, we moved the computer to another location and the colors on the screen changed markedly with the different voltage at the new electrical outlet. This change provides a reminder of the variation to which commercial monitors are subject. The measured specifications listed are a much more precise documentation of the test colors than the R G B numbers, which should not be trusted for precise replication of the test schemes. 175 176 Table E.l Experiment 2 Color Specifications Map 1: Soil-Drainage Classes for Forested Areas 1 DE Correct response for all trials is Different gray surrounds gray magenta dark magenta surrounds medium magenta dark magenta surrounds gray magenta and md magenta Induced trial: Controlled trial: Map Category Soil Drainage: very poor poor poor-imperfect imperfect imperfect-well (s: gray magenta) adjust by decreasing lightness: well drained Color Name 47 dk blue 55 gray blue 62 gray 55 gray-magenta 46 md magenta 40 md magenta adj 35 dk magenta 29 brown Hunt: RG -1.36 -0.01 1.16 2.34 3.57 3.70 5.47 2.38 (s: md magenta 8' c.) non-agricultural land YB -3.74 -2.06 -0.95 -0.94 -0.77 -0.61 0.15 1.10 Color Name dk blue gray blue gray gray magenta md magenta dk magenta brown I 46.59 55.48 62.43 55.10 46.15 40.14 35.17 29.06 cd/m2 L 10.7 16.3 21.5 15.3 9.74 7.09 5.12 3.71 In 100's R 63 335 284 437 437 398 344 347 403 437 323 213 171 77 220 CIE: x 0.174 0.226 0.275 0.298 0.326 0.332 0.404 0.382 B 437 437 437 376 293 245 182 1 1 0 0.172 0.219 0.246 0.230 0.218 0.217 0.223 0.312 Table E.1 (continued) Map 2: Land Use 2 QL 177 Correct response for all trials is Different Induced trial: light orange surrounds light cyan dark cyan surrounds medium cyan Controlled trial: Map Category residential commercial farm buildings livestock: fowl livestock: cattle other agricultural adjust by desaturating: recreational wildlife reserve Color Name 61 dk orange 73 1t orange 68 green 58 dk cyan 76 md cyan 80 1t cyan 79 1t cyan adj 85 gray 55 magenta (s: It cyan 8 control) (5: md cyan) Hunt: RG 2.94 1.76 -2.97 -2.46 -2.39 -1 .08 —0.48 -0.19 4.34 YB 3.20 2.85 2.77 0.02 0.30 0.49 0.14 -0.87 0.42 Color Name dk orange lt orange green dk cyan md cyan It cyan 81'3)’ magenta chm2 J L 61.09 19.2 72.72 30.4 68.50 28.0 58.36 19.0 75.68 36.6 79.95 42.0 79.21 41.0 84.45 48.0 54.61 14.2 light orange surrounds light and medium cyan In 100's R G 569 655 154 59 384 184 410 477 530 526 262 0.505 0.451 0.274 0.213 0.227 0.266 0.272 0.256 0.395 B 49 142 162 325 432 448 492 647 269 0.421 0.420 0.510 0.328 0.346 0.340 0.315 0.271 0.262 Table 13.1 (continued) Map 3: Correct response for all trials is Different light yellow-green surrounds dark orange medium red surrounds dark red light yellow-green surrounds dark orange and dark red Induced trial: Controlled trial: 178 Map Category Soil Type: Ag. peat gleyed and / or peat (s: dk or 8 c.) gleyed gleyed and/ or orthic orthic (s: dk red) Non-Ag. peat gleyed and / or peat gleyed gleyed and/or orthic adjust toward yellow: orthic Hunt: Color Name RG YB 29 md green -1.48 1. 94 57 1t yellow-green -0.14 1.94 67 gray 1.97 -0. 10 54 1t orange 4.11 1.44 36 md red 6.76 2. 23 16 dk green 0.25 1.03 45 dk yellow-green 0. 00 1.80 52 dl< gray 202 -0. 03 36 dk orange 4.99 1. 93 39 dk orange adj 5.04 2. 38 27 dk red 5.60 1. 78 Color Name Soil-Type Distributions for Agricultural versus Non-Agricultural Land Uses 2 DE/N md green lt yellow-green gray lt orange md red dk green dk yellow-green dk gray dk orange dk red chm2 J L 29.00 3.89 56.87 17. 3 67. 28 25.3 54.30 14.1 36.42 5.45 15.77 1.33 45.39 10.1 52.06 13.5 35. 54 5.32 38.99 6.55 27.47 3.13 In 100's R G 564 450 530 262 0 120 254 331 400 320 341 121 381 152 280 160 X 0.284 0.351 0.323 0.439 0.601 0.300 0.350 0.321 0.507 0. 542 0.573 B 161 412 206 11 120 293 90 62 0.534 0.427 0. 271 0.309 0.339 0.422 0.425 0.268 0.323 0.357 0.333 Table E.1 (continued) 179 Map 4: Soil-Drainage Classes for Agricultural Land 1 DE Correct response for all trials is Different Induced trial: Controlled trial: Map Category Soil Drainage: very poor poor poor-imperfect adjust toward orange: imperfect imperfect-well well drained non-agricultural land Color Name 61 dk orange 69 md orange 71 gray orange 71 gray orange adj 69 1t gray 6O gray purple 50 dk purple 35 dk gray Hunt: RG 2.15 1.41 0.46 0.66 0.19 1.05 2.04 0.35 (s: gray orange) (5: It gray 8 control) 3.16 2.73 1.09 1.55 -0.27 -1.25 -2.72 0.19 Color Name dl< orange md orange gray orange lt gray gray Purple dk purple dk gray I 60.56 68.62 70.87 71.08 68.96 60.12 49.73 35.28 dark orange surrounds gray orange dark purple surrounds light gray dark purple surrounds gray orange and light gray cd/m2 L 19.1 26.5 29.9 29.8 28.5 19.6 11.9 5.82 In 100's R G 535 302 588 386 513 441 545 433 437 437 406 325 345 195 175 175 X 0.487 0.437 0.336 0.362 0.274 0.264 0.244 0.271 9 138 262 437 437 437 ' 175 Y 0.441 0.425 0.357 0.379 0.287 0.236 0.172 0.290 180 Table 15.1 (continued) Map 5: Soil Drainage and Distance to Surface Water 2 SS/SS Correct response for all trials is Different Induced trial: medium cyan surrounds light cyan light magenta surrounds light purple Controlled trial: light magenta surrounds light cyan and light purple In 100's Map Category Color Name R C Soil Drainage: Distance to Surface Water in Kilometers: poor 0-1 lt purple 594 513 adjust by increasing lightness: 636 578 1-2 lt cyan 375 490 2-3 (s: It cyan) md cyan 157 466 imperfect 0—1 (s: It purple 8 control) It magenta 624 347 1-2 md purple 404 258 2-3 blue 185 235 well drained 0—1 md magenta 655 180 1-2 magenta-purple 434 92 2-3 purple 214 3 Hunt: cdlm2 CIE: Color Name RG YB J L x 76 1t purple 1.22 -1.32 76.42 35.6 0.270 81 1t purple adj 1.09 -1.09 80.92 41.5 0.274 681t cyan -0.11 -1.91 68.11 27.2 0.232 62 md cyan -1.69 -2.35 62.21 21.9 0.197 67 1t magenta 2.90 -0.71 67.34 24.6 0.322 55 md purple 2.17 -2.83 54.62 14.8 0.246 45 blue 0.25 -4.04 44.99 9.62 0.192 60 md magenta 5.29 0.27 59.79 17.2 0.409 46 magenta-purple 4.70 -1.84 45.92 9.27 0.313 33 purple 2.27 -4.71 33.46 4.77 0.202 B 620 655 570 521 470 ~ 454 320 354 0.234 0.244 0.229 0.228 0.237 0.170 0.148 0.248 0.173 0.104 Table E1 (continued) Map 6: Soils Series 1 QL 181 Correct response for all trials is Different Induced trial: Controlled trial: Map Category Red River Marquette 8: Kline Kline Osbourne adjust toward orange: Inwood 8: Meleb Berry Is. 8: Beaverdam Inwood 8: Beaverdam (5: yellow 8 control) Inwood Leary Gunton Aneda 8: Gunton Color Name 53 dk purple 66 pink 65 orange 88 yellow 84 yellow adj 74 cyan 58 blue 65 md purple 75 It purple 84 1t green 66 md green 56 dk green (5: It green) Hunt: RG YB 4.34 -3.27 3.86 -1.56 3.57 3.27 -0.89 3.49 -0.08 3. 01 -2.29 -0. 88 -0.88 -3. 45 1.00 -2. 71 0.94 -1.17 -1.97 3.67 -2.68 2.48 -3.34 3.28 light purple surrounds yellow dark green surrounds light green light purple surrounds yellow and light green Color Name In 100's R 494 655 646 610 655 151 188 422 548 441 171 G 67 223 269 655 579 543 322 330 438 655 482 dk purple pink orange yellow cyan blue md purple lt purple lt green md green dk green cdlm2 J L 52. 55 12.6 66.38 22. 9 64.57 21. 7 87.72 51.5 84.19 45. 9 73.92 34.4 57.54 17. 7 65.02 23.4 75.36 34.7 84.27 46.9 66.40 25.9 56.24 17.1 28 CIE: x 0.270 0.313 0.518 0.390 0.407 0.208 0.191 0.233 0.268 0.351 0. 274 0.279 383 513 525 54 60 119 527 547 607 . 590 51 179 43 0.140 0.198 0. 406 0.513 0.478 0.289 0.182 0.188 0.243 0.543 0.486 0.576 Table E.1 (continued) 182 Map 7: Agricultural Land Uses and Aquifer Recharge 1 QLIN Correct response for all trials is Different Induced trial: Controlled trial: Map Category Aquifer Recharge Zone: Inside hay crops pasture non-agricultural Outside hay adjust toward orange: crops pasture non-agricultural (s: gray y 8 c.) agricultural clearing (s: gray gn) in progress Color Name 84 yellow 75 orange 81 green 81 pink 71 gray yellow 70 gray yellow adj 63 gray orange 69 gray green 66 gray pink 70 gray blue Hunt: RG -0.49 1.13 -1.33 0.97 -0.40 0.13 1.10 -1.31 0.98 -1.54 3.36 3.17 3.08 -0.13 3.18 3.12 2.97 3.01 -0.05 -0.31 pink surrounds gray yellow gray blue surrounds gray green pink surrounds gray yellow and gray green Color Name yellow orange green pink gray yellow gray orange gray 81' 99“ gray pink gray blue I 83.74 74.80 81.15 80.81 70.86 69.94 62.63 68.84 66.01 70.18 cd/m2 L 45.4 33.0 42.5 41.7 29.6 28.4 21.2 27.9 24.9 30.1 621 655 474 655 In 100's 621 633 524 452 352 492 387 513 X 0.403 0.451 0.352 0.301 0.401 0.423 0.445 0.350 0.300 0.235 61 87 144 47 l 1 2 395 443 0.502 0.453 0.500 0.284 0.500 0.488 0.454 0.502 0.285 0.305 Table 15.] (continued) 183 Map 8: Agricultural Land Uses and Aquifer Recharge 2 QLIN Correct response for all trials is Different dark green surrounds dark yellow Induced trial: light gray surrounds light yellow Controlled trial: Map Category Aquifer Recharge Zone: Inside hay crops pasture non-agricultural Outside hay adjust by desaturating: crops pasture non-agricultural agricultural clearing in progress Color Name 76 dk yellow 69 dk yellow-green 64 dk green 61 dk gray 84 1t yellow 85 1t yellow adj 79 1t yellow-green 78 1t green 74 1t gray 78 pink (5: dk yellow) (5: It y 8 c.) Hunt: RG YB -0.95 2.52 -2.12 2.24 -2.84 1.94 0.10 -0.05 -0.56 1.72 -0.36 1.04 -1.09 1.40 -1.93 0.96 0.10 -0.18 0.90 0.83 Color Name dk yellow dk yellow-green dk green dk gray lt yellow lt yellow-green 1t green 1t gray pink cd/m2 J L 75.87 35.8 69.47 29.0 64.50 24.2 60.74 20.8 84.37 47.4 84.61 48.1 79.03 40.4 77.61 38.9 74.32 34.6 78.44 38.4 R 256 1 14 353 559 424 290 light gray surrounds dark yellow and light yellow In 100's G 530 495 461 353 612 607 577 478 X 0.344 0.286 0.250 0.275 0.328 0.308 0.294 0.255 0.276 0.339 189 210 220 353 317 - 405 338 391 0.464 0.460 0.451 0.295 0.412 0.369 0.395 0.377 0.293 0.334 184 Table 15.1 (continued) Map 9: Land Cover 1 QL Correct response for all trials is Different Induced trial: dark brown surrounds dark blue yellow surrounds light blue Controlled trial: dark brown surrounds both blues Map Category Color Name non-vegetated gray forested recreation site magenta mixedwood 1t brown hardwood (s: dk blue 8 control) dk brown willow and shrub dk blue adjust by desaturating: prarie and meadow lt blue cultivated hayland (s: It blue) yellow cultivated cropland lt green pasture dk green agricultural land being cleared purple Hunt: cd/m2 Color Name RG YB J L 57 gray 0.18 -0.09 57.08 17.8 60 magenta 4.26 0.27 60.50 18.2 53 1t brown 0.97 1. 62 52.96 14.3 30 dk brown 2.34 1. 79 29.95 3.94 45 dk blue -1.66 -2. 29 44. 92 10.0 47 dk blue adj -1.05 -1.38 46.54 10.9 60 1t blue -2.03 -2.58 59. 61 19.7 86 yellow -0.26 1.68 85.67 49. 1 69 1t green -1.95 3.54 69.10 28. 2 54 dk green -1.78 3.14 53.50 15.0 42 purple 2.39 -2.90 41.57 7.80 In 100's R 327 371 224 G 323 206 271 1 12 152 326 224 287 X 0. 274 0.388 0.370 0.489 0.187 0.215 0.186 0. 340 0.349 0.345 0.241 251 259 376 612 332 328 323 175 336 310 319 33 ' 371 0.291 0.259 0.377 0.429 0. 221 0. 250 0. 221 0.408 0.548 0. 551 0.157 Table E1 (continued) 185 Map 10: Soil-Drainage Classes for Forested Areas 2 SS Correct response for all trials is Different Induced trial: dark gray surrounds medium-light gray white surrounds light gray Controlled trial: Map Category Soil Drainage: very poor poor poor-imperfect imperfect (s: md-lt gray) adjust by decreasing lightness: imperfect-well well drained not forested Color Name 13 black 37 dk gray 51 md gray 70 Ind-1t gray 64 md-lt gray adj 831t gray 91 white 46 gray magenta Hunt: RG 0.94 0.29 0.28 0.14 0.18 0.12 0.12 1.64 (s: It gray 8 control) YB 0.81 0.20 -0.07 -0.25 -0.19 -0.31 -0.32 0.18 In 100's G R 0 175 437 381 655 314 white surrounds medium-light gray and light gray 0 175 437 381 655 204 Color Name black dk gray md gray md-lt gray lt gray white gray magenta cd/m2 J L 12.75 0.97 37.11 6.50 51.46 13.8 70.42 30.1 63.64 23.3 83.42 46.6 90.71 58.1 46.21 10.3 X 0.294 0.272 0.274 0.274 0.274 0.275 0.276 0.317 0 175 437 381 655‘ Y 0.320 0.293 0.288 0.289 0.289 0.289 0.289 0.279 186 Table E.1 (continued) Map 11: Land Use 1 QL Correct response for all trials is Same Induced trial: dark purple surrounds orange beige surrounds orange Controlled trial: beige surrounds both oranges Map Category Color Name residential lt purple commercial (5: orange) dk purple farm buildings red livestock: fowl orange livestock: cattle brown adjust by desaturating: other agricultural (5: orange 8 c.) beige recreational pink wildlife reserve green Hunt: Color Name RG YB J 66 1t purple 1.07 -3.09 66.00 39 dk purple 1.45 -4.31 39.37 43 red 6.68 2.28 42.99 69 orange 2.65 2.94 68.89 61 brown 2.41 2.37 60.86 65 brown adj 1.63 1.70 64.75 76 beige 0.71 2.39 75.96 64 pink 2.73 -0.07 64.20 66 green -2.75 0.62 66.39 cd/m2 L 24.1 6.93 7.80 26.0 19.3 23.0 34.9 22.0 26.2 R 215 453 655 530 530 32 In 100's G 322 122 326 332 459 282 474 X 0.27 0.201 0.550 0.477 0.445 0.393 0.403 0.341 0.221 655 121 149 188 376 331 ‘ 0.175 0.126 0.313 0.405 0.385 0.365 0.424 0.263 0.366 187 Table E.1 (continued) Map 12: Soil Drainage and Distance to Surface Water 1 88/88 Correct response for all trials is Same Induced trial: dark gray surrounds gray yellow light yellow surrounds gray yellow Controlled trial: dark gray surrounds both gray yellows In 100's Map Category Color Name R C Soil Drainage: Distance to Surface Water in Kilometers: poor 0-1 (s: gray yellow 8 control) dkgray 120 120 1-2 gray blue 146 190 2-3 md blue 177 283 imperfect 0-1 gray yellow 392 307 adjust by decreasing lightness: 343 269 1-2 md gray 347 347 2-3 It blue 375 428 well drained 0-1 md yellow 564 416 1-2 (s: gray yellow) lt yellow 569 495 2-3 It gray 573 573 Hunt: cd/m2 CIE: Color Name RG YB J L x 7 dk gray 0.44 0.65 27.11 3.38 0. 284 42 gray blue -0.11 -3.53 41.84 8. 26 0.191 56 md blue -0.75 -4.76 56.24 16. 4 0.176 56 gray yellow 0.56 2.76 56.40 16.6 0. 422 51 gray yellow adj 0.51 2.63 51 .40 13.3 0.418 60 md gray 0.09 0. 00 60.02 20. 2 0.276 71 1t blue -O.26 -2. 13 70.55 29. 7 0.226 71 md yellow 0.60 3.29 71.10 29. 3 0.446 77 1t yellow 0.25 1.52 77.15 37.1 0.351 841t gray 0.07 —0.20 83.56 46.9 0.277 120 655 69 347 614 17 295 573 0.316 0.163 0.141 0.466 0.468 0.297 0. 223 0.481 0.389 0.294 188 Table 5.1 (continued) Map 13: Sand 8: Gravel Resources 2 QL Correct response for all trials is Same Induced trial: blue surrounds dark green dark brown surrounds dark green Controlled trial: dark brown surrounds both dark greens In 100's Map Category Color Name R G B beach ridges red 530 148 158 littoral sand 8: gravel orange 655 276 226 glaciofluvial sand 8: gravel dk beige 582 357 272 glacial till lt beige 655 494 342 alluvium (s: dk green) blue 222 352 418 lacustrine clay blue-green 258 444 379 peat dk green 185 319 272 adjust by increasing saturation: 113 319 253 . no data gray 567 567 567 roads (5: dk green 8 control) dk brown 151 95 23 Hunt: cd/m2 CIF; Color Name RG YB I L x y 51 red 5.30 1.86 51.11 11.9 0.487 0.311 64 orange 3.93 1.73 63.65 20.8 0.449 0.326 65 dk beige 2.38 1.19 64.84 22.7 0.388 0.325 75 1t beige 1.53 1.08 74.88 33.4 0.365 0.335 54 blue -0.49 -2.09 53.92 15.3 0.217 0.223 61 blue-green -1.03 -0.45 60.76 20.9 0.241 0.291 48 dk green -0.82 -0.22 48.34 12.0 0.242 0.293 47 dk green adj -1.57 -0.01 46.91 11.2 0.226 0.311 78 gray 0.59 -0.62 78.40 39.0 0.276 0.269 24 dk brown 1.68 1.49 23.91 2.59 0.412 0.389 189 Table 5.1 (continued) Map 14: Soil-Drainage Classes for Agricultural Land 2 SS Correct response for all trials is Same Induced trial: light gray surrounds medium gray black surrounds medium gray Controlled trial: light gray surrounds both medium grays Map Category Soil Drainage: very poor poor (5: md gray 8 control) poor-imperfect imperfect adjust by decreasing lightness: imperfect-well adjust by decreasing lightness: well drained (5: md gray) non-agricultural land Hunt: Color Name RG YB 91 white 0.06 -0.27 831t gray 0.07 -0.23 70 md-lt gray 0.07 -0.15 54 md gray 0.14 -0.01 51 md gray adj 0.14 0.04 36 dk gray 0.25 0.30 31 dk gray adj 0.33 0.39 14 black 0 80 0.84 46 gray blue -0.69 -0.30 Color Name white It gray md-lt gray md gray dk gray black gray blue cd/m2 J L 90.75 58.3 82.94 46.0 69.80 29.5 54.34 15.8 51.07 13.6 36.28 6.19 31.43 4.56 14.02 1.11 46.24 10.8 In 100's 655 655 437 437 175 175 150 150 43 43 187 258 X 0.276 0.276 0.275 0.274 0.274 0.274 0.274 0.292 0.242 655 437 175 150 - 263 0.292 0.293 0.294 0.294 0.295 0.300 0.301 0.330 0.287 190 Table 5.1 (continued) Map 15: Soils Series 2 QL Correct response for all trials is Same Induced trial: dark orange surrounds medium purple light purple surrounds medium purple Controlled trial: light purple surrounds both medium purples In 100's Map Category Color Name R G Red River red 565 167 Marquette 8: Kline (5: ml purple) dk orange 608 292 Kline lt orange 655 380 Osbourne yellow-orange 655 520 Inwood 8: Meleb dk purple 302 58 Berry Is. 8: Beaverdam md purple 378 175 adjust by desaturating: 369 220 Inwood 8: Beaverdam gray purple 500 380 Inwood (s: md purple 8 control) lt purple 608 537 Leary cyan 27 411 Gunton 1t blue 305 515 Aneda 8: Gunton dk blue 142 311 Hunt: cd/m2 CIE: Color Name RG YB J L x 54 red 5.21 3.02 53.78 13.5 0.552 63 dk orange 3.18 2.95 62.99 20.6 0.492 70 1t orange 2.24 2.08 70.36 27.8 0.427 78 yellow—orange 0.81 1.96 78.45 38.0 0.387 43 dk purple 2.48 -5.05 42.79 8.04 0.207 54 md purple 1.86 -4.74 54.02 14.1 0.208 54 md purple adj 1.74 -4.00 54.45 14.6 0.218 68 gray purple 1.37 -1.90 68.21 26.4 0.257 81 1t purple 0.86 -1.15 81.09 42.0 0.268 62 cyan -1.80 -3.90 62.39 21.7 0.174 72 1t blue -1.13 -1.81 71.65 31.3 0.219 52 dk blue -1.05 -3.00 51.76 13.8 0.192 47 101 217 255 517 655 590 580 ‘ 655 655 582 448 Y 0.373 0.398 0.372 0.398 0.105 0.121 0.141 0.211 0.245 0.176 0.245 0.196 Table 5.1 (continued) Map 16: Land Cover 2 QL 191 Correct response for all trials is Same Induced trial: blue surrounds orange yellow surrounds orange Controlled trial: Map Category non-vegetated forested recreation site mixedwood adjust by desaturating: hardwood willow and shrub prarie and meadow cultivated hayland cultivated cropland pasture agricultural land being cleared Color Name 39 gray 60 pink 53 brown 51 brown adj 65 orange 56 blue 73 cyan 83 yellow 78 green 45 red 54 purple (3: orange 8 c.) (5: orange) Hunt: RG -0.23 4.78 3.53 2.77 3.28 -0.88 -2.19 -0.45 -2.92 5.25 3.23 YB 0.01 -1.76 2.47 1.81 3.33 -4.01 -1.02 3.38 2.66 0.61 -4.52 blue surrounds both oranges Color Name gray pink brown orange blue cyan yellow green red purple I 39.09 60.04 53.36 51.11 65.37 56. 22 72. 59 82.83 77.79 44.77 54.32 cd/m2 L 7.41 17.5 13.7 12.6 22.5 16. 6 32.8 44.1 38.8 8. 75 13. 8 In 100's R 206 655 497 439 655 173 171 61 1 214 G 206 115 218 320 556 616 655 470 0.277 0.337 0.480 0. 427 0.515 0.183 0. 208 0. 406 0. 274 0.422 0.229 206 476 125 162 529 217 212 655 0.291 0.185 0.371 0.350 0.415 0.164 0.282 0.503 0. 499 0.254 0.117 192 Table 5.1 (continued) Map 17: Soil-Type Distributions for Agricultural versus Non-Agricultural Land Uses 1 SS/N Correct response for all trials is Same Induced trial: dark blue surrounds medium blue-green gray surrounds medium blue-geen Controlled trial: gray surrounds both medium blue-geens In 100's Map Category Color Name R G B Soil Type: Ag. peat md blue 0 265 383 gleyed and/or peat md blue-geen 114 371 347 adjust by increasing lightness: 126 413 385 gleyed md green 126 543 379 gleyed and/ or orthic 1t green 375 599 503‘ orthic white 575 655 608 N on-Ag. peat (s: md blue-green) dk blue 17 151 211 gleyed and/or peat dk blue-geen 67 256 238 gleyed dk green 117 362 266 gleyed and/ or orthic gray green 240 398 328 orthic (s: md blue-green 8 c.) gay 466 526 486 Hunt: cd/m2 CIE: Color Name RG YB J L x y 47 md blue -1.74 -2.92 46. 91 11.0 0.179 0.202 56 md blue-geen -2.11 -0.65 56.22 17.3 0.209 0.292 61 md blue-geen adj -2.18 -0.64 60. 69 20.9 0. 210 0.295 71 md green -2.68 0.53 71.38 31.5 0. 223 0.361 80 1t geen -1.39 0. 00 79.58 41. 6 0.247 0. 319 87 white -0.24 -0. 26 87. 49 53.1 0. 270 0. 296 30 dk blue -0.84 -1.76 29.67 4.07 0.187 0. 211 43 dk blue-geen -1.82 -0. 27 43.22 9.26 0.210 0.299 55 dk geen -2.23 0. 55 54.87 16.3 0.227 0. 352 60 gay geen -1.34 0.19 60.37 20.6 0.246 0.323 76 gay -0.30 -0.06 76.30 37.2 0.271 0.303 193 Table 5.1 (continued) Map 18: Sand 8: Gravel Resources 1 QL Correct response for all trials is Same Induced trial: light geen surrounds dark beige dark geen surrounds dark beige Controlled trial: dark geen surrounds both dark beiges In 100's Map Category Color Name R G B beach ridges lt beige 651 545 349 littoral sand 8: gavel dk beige 522 437 280 adjust by decreasing lightness: 487 408 261 glaciofluvial sand 8: gravel lt cyan 313 552 477 glacial till dk cyan 222 392 339 alluvium (s: dk beige) 1t geen 377 625 351 lacustrine clay md geen 263 435 244 peat (s: dk beige 8 control) dk geen 153 306 153 no data It gay 409 398 397 roads dk gay 216 216 216 Hunt: cd/m2 CIE: Color Name RG YB J L x y 84 1t beige 0.42 1.32 83.67 45.7 0.349 0.373 72 dk beige 0.41 1.38 71.88 30.9 0.348 0.376 69 dk beige adj 0.42 1.39 68.95 27.8 0.348 0.376 771t cyan -1.58 -0.04 76.84 38.0 0.241 0.319 61 dk cyan -1.50 0.12 61.29 21.4 0.241 0.322 83 1t geen -1.65 1.47 82.54 45.4 0.279 0.406 65 md geen -1.61 1.59 65.12 24.7 0.279 0.411 50 dk geen -1.86 1.92 49.99 12.9 0.274 0.443 66 1t gay 0.18 -0.06 66.47 26.0 0.279 0.295 43 dk gay 0.17 0.19 42.84 8.99 0.274 0.299 194 Table 5.1 (continued) Map 19: Evaluation of Potential Campground Sites 1 D5 Correct response for all trials is Same Induced trial: pink surrounds dark magenta dark brown surrounds dark magenta Controlled trial: dark brown surrounds both dark magentas In 100' Map Category Color Name R G Ranking of Sites: 1 optimal dk red 272 58 adjust by increasing saturation: 272 5 2 dk magenta 366 116 3 md magenta 426 212 4 gay magenta 437 280 5 (s: dk magenta) pink 487 396 6 minimally acceptable It gay 480 480 7 gay brown 437 388 8 md brown 345 255 9 unacceptable (s: dk magenta 8 c.) dk brown 246 182 roads black brown 125 92 Hunt: cd/m2 CIE: Color Name RG YB J L x 31 dk red 4.39 0. 86 30.55 3.93 0.420 29 dk red adj 5.05 1. 52 28.61 3.41 0.494 41 dk magenta 4.00 0.55 40. 99 7.39 0.391 51 md magenta 2.71 -0.12 51.49 12.9 0.334 58 gay magenta 1.72 .039 58.04 17.7 0.303 69 pink 0.80 -0.23 68.80 27.8 0.290 741t gay 0.08 -0.16 74. 22 34.5 0. 276 65 gay brown 0.39 0.53 65.46 24.7 0. 308 50 md brown 0.99 1.70 50.42 12.7 0. 375 39 dk brown 0.94 1.70 39. 24 7.15 0.375 22 black brown 1.08 1.31 22.13 2.28 0.366 127 87 191 359 434 335 154 110 0.263 0.295 0. 263 0. 256 0. 259 0.280 0. 294 0. 324 0.383 0.392 0.385 Table 5.1 (continued) 195 Map 20: Evaluation of Potential Campground Sites 2 DE Correct response for all trials is Same Induced trial: Controlled trial: Map Category Ranking of Sites: 1 optimal \lmUIu-FOJN minimally acceptable (5: green-y) adjust by increasing lightness: 8 9 unacceptable roads Color Name 39 dk red 50 md red—orange 61 1t red-orange 71 1t orange 77 orange-yellow 84 yellow 66 green-yellow 70 geen-yellow adj 53 md geen 38 dk geen 28 brown Hunt: RG 6.64 5.20 3.20 1.87 0.89 0.27 0.10 0.07 -0.29 -0.87 1.58 (s: green-y 8 c.) YB 2.31 2.93 3.02 2.70 1.92 1.28 1.42 1.39 1.71 2.02 1.74 yellow surrounds geen-yellow dark geen surrounds geen-yellow dark geen surrounds both geen-yellows Color Name dk red md red-orange 1t red-orange R 409 521 588 648 650 652 431 479 291 141 185 In 100's G 16 169 303 435 545 655 517 359 241 131 X It orange orange-yellow yellow geen-yellow md green dk geen brown cd/m2 J L 38.62 6.20 50.39 11.6 60.83 18.9 71.02 28.6 77.35 36.6 83.81 46.1 65.51 24.6 69.90 29.0 53.21 14.8 37.64 6.68 27.94 3.48 0.600 0.563 0.498 0.447 0.387 0.343 0.339 0.338 0.331 0.316 0.440 12 34 96 158 260 361 242 269 167 87 33 0.347 0.381 0.403 0.41 1 0.394 0.375 0.387 0.386 0.413 0.468 0.429