- :u Ell|llllll LIBRARY Michigan State University This is to certify that the thesis entitled An Evaluated Set of Dithered Patterns For CRT Maps presented by Ann Marie Goulette has been accepted towards fulfillment of the requirements for M. A. Geography degree in Maj r professor Date May 13, 1987 0.7539 MS U i: an Affirmative Action/Equal Opportunity Institution )VIESI_J RETURNING MATERIALS: Place in book drop to LJBRARJES remove this checkout from Jun-lyilll. your record. FINES will be charged if book is returned after the date stamped below. AN EVALUATED SET OF DITHERED PATTERNS FOR CRT MAPS By Ann Marie Goulette A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Geography 1987 ABSTRACT AN EVALUATED SET OF DITHERED PATTERNS FOR CRT MAPS By Ann Marie Goulette This research examined a set of patterns created by the dither- ing technique for use as area symbols on maps displayed on a CRT monitor. Excluding patterns with very coarse texture or disturbing optical effects, a set of twenty-three patterns was initially chosen and subjected to further examination. The evaluation of these patterns focused on value estimation and pattern differen- tiability. The results indicated that, relative to the percentage of light-colored pixels, subjects underestimated the value of lighter toned patterns and overestimated the value of darker toned patterns. In addition, subjects could easily differentiate twelve patterns from the original twenty-three. The twelve patterns were used as area fill on choropleth maps using five different numbers of classes and seven color combinations. Subjects were asked to match patterns on the map with those in the legend. The maps containing the fewest classes and those with color schemes using dark colors dithered with white had the greatest number of correct matches. Five patterns, two of them solid colors, were correctly matched 95% of the time or better. It was concluded that if a large palette of solid colors is unavailable, dithered patterns can be used effective- ly for maps requiring a small number of distinguishable patterns. ACKNOWLEDGEMENTS I want to thank my advisor, Dr. Judy M. Olson, and Dr. Richard E. Groop for their help throughout my academic career and their guidance on this thesis. In addition, Mr. J. Michael Lipsey had a great influence in my transformation into a cartographer. I am grateful also to the Geography Department of Michigan State Univer- sity for their efforts to help me think geographically and for their generous grants and student assistantships. I also thank my friends and co-workers for their subtle (and not so subtle) encour~ agement to complete this thesis. Finally, I thank my parents, Van and Mary Goulette, for their love and care. ii LIST OF TABLES . . . LIST OF FIGURES . . CHAPTER I. II. III. INTRODUCTION . Historical Context of Cartographic Area Symbols TABLE OF CONTENTS Produced by Computers . . . The Cathode Ray Tube (CRT) Monitor . . . Dithering as a Proposed Improvement for Cartographic Area Symbols on CRT Monitors . Research on CRT Color Displays . The Problem Structure of the Thesis . . . PRELIMINARY STEPS: DEVELOPING TEST Equipment . Software . The Patterns The Slides . EVALUATION, PART I: VALUE TIATION, PREFERENCE . Method . . Test Slides Subjects Testing Environment Procedure ESTIMATION, iii MATERIALS PATTERN DIFFEREN- Page vi 10 IO 12 12 13 15 24 27 27 28 3O 3O RESUItS O O O O O O O 0 O O O O O 0 Discussion . . . . . . . . . . . . . IV. EVALUATION, PART II: DITHERED PATTERNS IN MAP CONTEXT Method . . . . . . . . . . . . . . Test Slides . . . . . . . . . . . Subjects . . . . . . . . . . . . Testing Environment . . . . . . . . . Procedure . . . . . . . . . . . . Results . . . . . . . . . . . . . . Discussion . . . . . . . . . . . . . V. SUMMARY AND CONCLUSIONS . . . . . . . . . APPENDIX A. BASIC Computer Program to Create Dithered Patterns B. BASIC Computer Program to Create a Choropleth Map with Dithered Patterns . . . . . . . . . C. Consent Form . . . . . . . . . . . . . D. Test Booklet and Instructions . . . . . . . E. Data Values and Classes . . . . . . . . . F. Slide Presentation Order in Study 2 . . . . . BIBLIOGRAPHY O O O O O O O O O O O O O O 0 iv 31 36 42 42 44 44 44 45 57 6O 63 68 69 71 73 75 LIST OF TABLES Page Distribution of Foreground and Background Pixels in Preliminary Dithered Patterns . . . . . 19 Distribution of Foreground and Background Pixels in Dithered Patterns Chosen for Study . . . 25 Pattern Pairs Chosen for Study . . . . . . . 29 Estimates of Value Percentages of Royal Blue/White Patterns 0 O O O O O O O O O O O O 32 Estimates of Value Percentages for Different Colored Patterns . . . . . . . . . . 32 Perceived Amount of Differentiability Between Royal Blue/White Patterns and Those of Different Color Schemes . . . . . . . . . . . 35 Number of Correct Responses to Pattern Pairs . . 37 Pattern Pairs Judged to Be "Same" or "Different" . 38 Mean Correct Responses Per Pattern . . . . . 48 Mean Correct Responses Per Number of Map Classes . 48 Pattern Differentiability . . . . . . . . 50 Mean Correct Responses Per Color Scheme . . . . 51 Analysis of Variance of Color Scheme Treatments . 53 53 Mean Correct Responses Per Dither Color Type . Analysis of Variance of Dither Color Type Treatments . . . . . . . . . . . . 53 Figure Examples of Light-Background and Dark-Background Dithered Patterns Using a Triangular Lattice Arrangement of Pixels Dithered Patterns with Disturbing Optical Effects Redesigned Patterns with Middle-Range Values LIST OF FIGURES Test Patterns for Study 1 . Relationship Between Perceived and Actual Value Percentage of Royal Blue/White Patterns Examples of Patterns Judged to Be Differentiable in Study 1 . Counties Tested Number of Correct Responses for Maps with Different Numbers of Classes . Examples of Patterns Judged to Be Differentiable in Study 2 . vi Page 14 21 22 23 33 41 46 49 55 CHAPTER 1 INTRODUCTION Maps have been produced by computers for several decades, but only recently have cartographers begun to display maps on cathode ray tube (CRT) monitors. Recent decreases in price for this technology have made the monitors available to almost every cartographer, but little is known about the limitations of the device for cartographic output. As increas- ing numbers of maps are made on CRT monitors, solutions must be sought to improve CRT map design. Perhaps foremost in the cartographic design limitations of inex- pensive CRT monitors is their limited color palette. Most common mon- itors can display only 4, 8, or 16 colors. These color restrictions con- strain the cartographer in designing complex or aesthetically-pleasing displays. In particular, the use of cartographic area symbols is severe- ly constrained by narrow color choice because a logical progression of several tones is unavailable. The 4, 8, or 16 colors available on inex- pensive CRTs are highly contrasting and very bright when covering large areas on the monitor. Dithered patterns, commonly used in other computer graphics appli- cations, offer a solution to the limited area symbols available on CRT monitors. The dithering technique uses two or more of the colors avail- able on the CRT to create a pattern. The impression of a greater number of hues is achieved by the viewer's optical mixing of the pattern colors 1 2 into another hue not in the CRT palette. There are literally millions of possible dithered patterns, but to use dithered patterns successfully in creating CRT maps, cartographers need sets of patterns that are distinguishable from one another and that form a logical progression from light to dark. The purpose of this thesis is to develop a set of dithered patterns and evaluate their suitability for use as cartographic area symbols. Historical Context of Cartographic Area Symbols Produced by Computers The search for suitable cartographic area symbols from computer graphics output devices is not a novel idea. As each new technological advance made its way into the marketplace, cartographers were ready to use the devices for mapping applications. In general, early research on area symbols on a particular device yielded unsatisfactory results; further research focused on improving the output given the constraints of the device. In the late 1960's computer maps were made with line printers, the only available output device at the time, using mapping packages such as SYMAP. A milestone in computer mapping software, SYMAP was written to produce choropleth, isoline, or proximal maps on the line printer. Area symbols for these maps consisted of repetitive printing of elements of the character set of the printer. The output was coarse and blocky and offended the aesthetic sense of many cartographers (Carter, 1984, p.44). Refinements to line printer map output are documented in the lit- erature. Smith (1980) has found that SYMAP output can be improved by 3 selecting alternative symbols that enhance the gray tone contrast rather than accepting the default symbols. In some instances where line printers were used for making maps, special modifications were made to line printer characters to produce better cartographic symbols (Brassel, 1974). Many users also improved the image by photographically reducing output, thus making it appear less blocky. An important advance in computer hardware for mapping was the develOpment of pen plotters. This device enabled the cartographer to plot lines or vectors that resembled hand-drawn lines and offered a choice of pen widths and colors. In many cartographic applications, particularly in the display of base map material, plotter-drawn lines were an aesthetic improvement over line printer symbols. However, drawing lines did not answer the problem of creating attractive area symbols. Area fill patterns produced by such packages as CALFORM and SAS/GRAPH are limited to those created by drawing stripes or cross-hatching in various orientations. Various densities of patterns are created by varying the spacing of the lines. Nhen few lines are used, the patterns appear coarse; when many lines are specified, plotting time is considerable and so much ink is used that the paper can tear easily. The invention of the microchip and the proliferation of the micro- computer in the last decade has brought another output device into the hands of most cartographers: the dot-matrix printer. The dot-matrix printer, like the line printer, forms characters by a succession of firings of vertical pins (often nine vertical pins fired six times). The advantage of the dot-matrix printer over the line printer is that software can be written to fire the pins individually. This allows the creation 4 of a wide range of patterns. Recent cartographic research focusing on the use of the dot-matrix printer for creating pattern fills for mapping includes the works of Groop and Smith (1982) and Plumb and Slocum (1986). Both describe improvements in area symbolization for dot-matrix printer maps. Groop and Smith created regularly-spaced pattern fills; Plumb and Slocum produced fill patterns using a random-dot approach. The Cathode Ray_Tube (CRT) Monitor Like the dot-matrix printer, the cathode ray tube monitor was also made available through the proliferation of the microcomputer. The CRT is available in alphanumeric mode (which is of little use to cartography), graphics mode, or both. The monitors also come with monochrome or color capabilities. Regardless of these differences, however, the CRT display image is created by a scanning electron beam directed at the monitor's screen. Two factors affect the ability to create area symbols on CRTs: the resolution of the screen and the number of available colors. Resolution refers to the smallest readable character that can be displayed and the minumum spacing (vertical and horizontal) that can be discerned (Machover, 1977). The resolution of CRT monitors is measured in picture elements, or pixels. Expensive monitors often have higher resolution than less expensive models. For example, a Tektronix 41158 monitor has a resolution of 1280 x 1024 pixels; an IBM-PC microcomputer CRT has a resolution of 320 x 200 pixels in the color graphics mode. Even on the most expensive monitors, individual pixels are evident on CRT images, giving a jagged or stair-step effect. Some sophisticated software for expensive monitors employs a technique called "anti-alaising,“ 5 which utilizes varying colors or intensities of adjacent pixels to lessen the jagged appearance of lines and highly contrasting colored areas. The inexpensive monitors have neither sufficient resolution nor the color capacity to perform this technique. CRT hardware devices also vary greatly in the number of colors that can be displayed. Inexpensive monitors available today can display 4, 8, or 16 colors. More expensive monitors can display 256 colors or more, chosen from palettes in the thousands to millions. The variation in the number of colors and the number of colors available simultaneously is not due so much to the hardware, but to the amount of memory allocated for this purpose. Gray-scale (intensity) and color can require from 3 to 12 bits of memory per pixel (Hobbs, 1981). Monitors having high resolu— tion and a large number of available colors require large amounts of mem- ory that are generally not available in inexpensive systems. This lack of a full palette of colors severely limits the complex- ity of maps that a cartographer can make with inexpensive CRTs. Using an IBM-PC color monitor with a simple color board, for example, one must choose one of two four-color schemes. One color scheme offers black, yellow, red, and green; the other consists of black, white, cyan, and magenta. Using these colors as areal fills, the cartographer can produce only a 3-class choropleth map if the linework and lettering are highlighted in a color of their own. Cartographers frequently make maps requiring considerably greater complexity than this. In addition to the problem of few colors available, colors such as black, yellow, red, and green in combination are unattractive. On the most common monitors, the colors are of high saturation or chroma, which few cartographers find appealing when used in combination with one another. 6 The cartographer's options for CRT displays are limited not only by hardware, but also by software and programming languages available for the microcomputer. Both MS-BASIC and TURBO PASCAL support graphics capability on CRT monitors. But neither allows the programmer to draw lines or fill areas (using the convenient line drawing and area fill commands) with anything but solid colors. Clearly, if color CRT monitors are to be more useful for cartography, new solutions for areal symboliza- tion must be found through Specialized computer graphics programs. Dithering as a Proposed Improvement for Cartographic Area Symbols on CRT Monitors One solution for improved area symbols on CRT displays may be in the computer graphics technique called “dithering“ or halftoning. Dithered patterns are created by interspersing pixels of different colors or intensities. Dithering can refer to pixels arranged into repetitive patterns, to arrangements of randomly mixed pixels of different colors, or to pixels arranged to appear to grade from one color into another. Dithering has been used previously in several computer graphics applications. It has been used typically to display three-dimensional graphics and solid modeling on computer monitors (Ryan, 1983) and in computer art and animation. Dithered patterns are currently available for use on some microcom- puters, such as the Apple Macintosh and the Tektronix 41158. The Apple Macintosh uses dithering in its MacDraw and MacPaint software applications. Graphics software for the Tektronix 41158 has several dithered and hatched patterns available for use. In addition, both the Macintosh and the 7 Tektronix 41158 software allow the user to define new patterns. Despite this availability of dithered patterns, a survey of recent mapping software reveals little use of the patterns. Jensen (1986) notes that MICROPIPS Version 1.0, an image processing package written for the IBM-PC, makes it "difficult to select a group of colors using the standard IBM color palette that will give the impression of a graded progression from low to high brightness values.“ Jensen views MICROPIPS' inability to display shades of grey and the IBM's limited palette of 16 colors and low resolution as serious disadvantages for digital image processing. These disadvantages restrict the usefulness of the software for its in- tended purpose. Several other reviewers note the lack of patterned area symbols in graphics software. Groop (1985) reviewed the Desktop Information Display System (DIDS) Version 2.0 using an IBM-XT with a standard 4-color card. He notes that the "maps displayed were solid-filled areas rather than IBM 'tiling' [dithered] patterns, thereby limiting chor0pleth symbols to three distinct classes.” Similarly, the program PRODESIGN II, a computer-aided design package, does not support line, dot or hatch patterns to fill areas (Gossette, 1986). At least three software systems employ patterns to fill areas. AutoCAD has several hatch patterns to fill any shape (Morrison and Men- doza, 1986), as does Atlas AMP (Foote, 1985). These hatch patterns, however, are not true dithered patterns. The patterns consist of cross- hatched lines, a throw-back to the pen plotter hatch fills. The ODYSSEY system, created by the Harvard Laboratory for Computer Graphics and Spatial Analysis, does use a true dither. The POLYPS component of the system allows the user to choose dots, lines, or 8 patterns to fill areas (White, 1980). The patterns consist of regular- ly spaced light-colored pixels on a background of dark-colored pixels. Research on CRT Color Displays Even though CRT images are gaining acceptance and popularity, only a few studies have been conducted on the perceptual aspects of the mon- itor displays. The work of Haber and Wilkinson (1982) offers practical advise based on color theory in the design of CRT images. They suggest that programmers carefully choose colors, intensities, textures, and shadings to maximize contrast between adjacent features on the display. The study does not produce empirical evidence for their claims of improve- ment in CRT image design. Indeed, Haber and Wilkinson admit that the lack of research in this area makes design choices "more of an art than a science." Another descriptive study focuses on the use of pattern on CRT monitors. Truckenbrod (1981), an artist and computer graphics designer, has employed the pointillist theory of the French Impressionist painters in her study of effective use of color on CRT monitors. Using three sets of two-color dithered patterns, she has shown how the technique can effec- tively change the appearance of hue, value, and chroma. She has varied the percentages of two colors of pixels in inverse proportions to one another from 0% to 100%. She states that the apparent changes in hue, value, and chroma are the result of the viewer's optical mixing of the two colors, perceiving a color different than the two that it comprises. While Truckenbrod's patterns do effectively vary pixel color den- sities, they can be described as coarse or "busy." Many of the patterns have definite repetitive shapes within the overall pattern, creating the 9 appearance of rosettes, shells, and leaves. Although textural patterns are often appropriate in general graphic usage, these obvious sub-elements make the patterns unsuitable for mapping. Murch (1984) describes ways that color can be used effectively on CRT images. He notes that hue information is lost for small areas on the CRT and that not all colors are equally legible or readable. In contrast to his mostly intuitive suggestions, he quantified the amount of luminance (light emitted) for eight colors displayed on a CRT. Luminance values (cd/mz) range from 0 for black to 10 for white. Examples of intermediate values are 7.6 for yellow, 4.7 for red, and 2.7 for blue. These measures have implications for how CRT colors may be used and perceived when viewed in combination. Small areas of any color will be difficult to perceive. In addition, the contrast of luminance between areas of different colors on the CRT may affect color perception. Taylor (1983) points out a woeful lack of cartographic research on the CRT. He argues that "as cartographers we must not simply take the approach of modifying our existing maps to display them on [CRTs] -- we must also seek new solutions utilizing the strengths of the new medium.“ He cautions against the easy misuse of color on the monitors and calls for more cartographic research on color. Basic research on CRT maps, he says, will help the cartographic designer make better maps. McGranaghan is one of the few cartographers working on CRT map design. In an article written in 1985, he investigated the effect of CRT background color on the map reader's impression of "dark is more" on choropleth maps. He concluded that dark is more when the CRT background is light; conversely, more than 50% of his subjects felt that “light is more“ against a dark background (McGranaghan, 1985). He has 10 also evaluated the applicability of traditional color use guidelines to maps produced on CRT monitors. Based on a reaction-time experiment, he concluded that maps that vary symbols by saturation require more time to evaluate than maps that vary by value only (McGranaghan, 1986). In both studies, McGranaghan used only solidcolor area symbols. The Problem The general lack of research on CRT monitors and cartographic use of the device and the unsatisfactory nature of solid-color fill as cartographic area symbols on inexpensive devices have led to the research reported in this thesis. Because so little is known about the device and the patterns the inexpensive but ubiquitous CRT can produce, the goals of this research were to create a reasonable set of dithered patterns on an 8-color monitor and to evaluate them within the context of use on CRT choropleth maps. Two overriding questions were involved: how many patterns would be available (i.e., how many in a light-to-dark progression would be distinguishable), and how effective would they be when used in a map context. Structure of the Thesis Chapter 1 has included a brief look at the historical context of the problem, stressing the adjustments that have taken place as comput- er technology used in mapping has changed. It has also reviewed the literature most closely related to the thesis topic and has stated the thesis problem. 11 Chapter 2 covers the several basic steps that had to be taken to create the dithered patterns for this research. Computer software de- veTOpment, initial selection of patterns, and development of a hardcopy procedure for testing the patterns will be discussed. A Chapter 3 presents tests concerned with general perceptual parameters of the dithered patterns. In particular, value estimation, pattern differentiation, and color preference are investigated. Subjects were asked to estimate value percentage of patterns, distinguish differences between two patterns and choose a color scheme preference from seven color schemes. The results of this preliminary testing are intended to further narrow the range of dithered patterns appropriate for mapping. Chapter 4 presents the test evaluating the patterns in a cartographic context. The patterns are used in combination with one another on classed choropleth maps to investigate which patterns are most easily differenti- ated and which color schemes are most preferred. Color schemes are also investigated for differences in correct pattern matching responses. Chapter 5 summarizes the research and presents conclusions. The evaluated set of dithered patterns as cartographic area symbols is dis- cussed, and suggestions are given for potential cartographic applications of the patterns and for possible future research to refine them. CHAPTER 2 PRELIMINARY STEPS: DEVELOPING TEST MATERIALS The nature of this research necessitates that several steps must precede the experimentation. Software must be written to display the dithered patterns. Patterns of potential use for cartographic purposes must be chosen from the numerous ones possible, and the component colors must be selected. Also, an appropriate method to record the CRT images for testing purposes must be defined. This chapter describes the work conducted prior to testing dithered patterns with subjects. Eguipment Because computer equipment and especially color CRT monitors, differ widely in their capabilities, it is necessary to describe in detail the properties of the equipment used throughout this study. The display of patterns and analysis of data was accomplished on a Texas Instruments Professional Computer (TIPC) with 192K random access memory, a color adapter, and Texas Instruments color monitor. The color monitor features an 8-color palette; all eight colors can be displayed simultaneously on the screen. The resolution of the monitor is 300 pixels vertically by 720 pixels horizontally. The screen measures 13 inches diagonally. The operating system for the computer was MS-DOS, Version 1.10. Software was written using the MS-BASIC, Version 1.10 programming language. 12 13 Software Two original programs were required for this study. One program was written to do basic design work on dithered patterns. The other program was needed to display dithered patterns in a cartographic context. The program for designing the various dithered patterns is found in Appendix A. In essence, the program will display a square containing a dithered pattern using a triangular lattice arrangement of two colors of pixels from the monitor's palette. Figure 1 contains a black and white simulation of the display's image. The program prompts the user to enter the number of the chosen foreground and background colors and the hori- zontal and vertical increments for the arrangement of the foreground pixels. The program then fills a square on the screen with the background color and continues by coloring the pixels with the foreground color in the increments specified. Any color combinations and increments (up to the size of the displayed square) can be specified to create a large number of patterns. The program to fill a base map with dithered patterns is more com- plex. This program had to fill irregular polygons with the pre-specified patterns. Therefore, a polygon-fill algorithm had to be merged with the dithered pattern algorithm used in the pattern development work. The resultant algorithm was required to fill the polygons pixel by pixel to create the patterns. The software written to fill polygons with the dithered patterns was based on an algorithm by Pavlidis (1979, 1982). Appendix 8 contains the programming code to create a 4-class choropleth map. Within a pre-established maximum-minimum x-y pixel window for each polygon, the algorithm checks the color of each pixel inside the window against the Figure 1. Simulated examples of light-background and dark-background dithered patterns using a triangular lattice arrangement of pixels. This figure was oonstmcted using an Apple Macintosh computer and LaserWriter printer. 15 color of the polygon's boundary pixels. Moving horizontally across a tier of pixels, a flag is set to "even" until a pixel of the boundary color is encountered. At that point, the flag is set to “odd“ and the pixels above and below the pixel are checked to determine if the original pixel is a minimum or maximum point of the polygon. If the pixel is in- side the polygon, it is colored appropriately for the selected pattern, as are the other interior pixels in the horizontal scan. When another boundary pixel is encountered, the flag switches back to “even.“ The procedure continues until every pixel in the polygon's window has been processed. There are several disadvantages to the algorithm. In his articles, Pavlidis noted that raster-based polygon filling routines are "non-trivial" and often do not work perfectly in practice. Some complex polygons do not fill correctly and require decomposition into simpler shapes. This decomposition creates a greater number of polygons to fill and adds processing time to an already time-consuming algorithm. In addition, the algorithm works best in a recursive language like Pascal or Forth and is therefore complicated to write in Basic. Nevertheless, it fulfilled the needs of the work being performed here and the Basic language was used. The Patterns The use of patterns as cartographic symbols is certainly not new. The literature of the past 25 years is replete with studies involving the use of patterns on maps. Because no comparable literature is available to guide the choice of patterns for mapping on CRTs, a brief review of literature on the perception and preferences of subjects for patterns is presented here as background to the choice of patterns in this research. 16 The perception of several pattern variables has been studied by many researchers. Williams (1958), for example, sought a psychophysical function to describe the differences between the actual percentage of area inked on a pattern versus the subject's estimate of the darkness of the area. Williams found that the relationship was non-linear; subjects overestimated the lighter values and underestimated the darker values. Castner and Robinson (1969) identified six basic characteristics of dot patterns: 1) the form of the dots; 2) the size of the dots; 3) the linear distance between the dots; 4) the arrangement of dots; 5) the orientation of the arrangement; and 6) the pattern-value of the dot area symbol. The sixth characteristic, pattern-value, refers to the "total impression of gray value which results from the visual integra- tion of the form, size, spacing and arrangement of the dots and from the orientation of the arrangement.“ The authors stated that the per- ception of the visual pattern of the dots and the gray value of the pat- tern is greatly dependent upon the textural characteristics of the pattern. As dot spacing increases, the texture or arrangement of the dots them- selves, rather than value, is the first thing a map reader perceives. They added that patterns with fewer than 40 lines per inch will be per- ceived first as patterns of dots and "only with difficulty can a gray tone be perceived." Dent (1985, p. 213) discusses whether patterns need to be differenti- able on a map. The school of thought that views the map as an areal table carries the idea that each pattern must be visually discernable from its neighboring patterns. The other school concentrates on the over- all geographical distribution of data on the map. To them, pattern differentiability is not as important. 17 Jenks and Knos (1961) found that map users prefer dot patterns to linear or irregular patterns in a graded series. In addition, user preferences tended toward uniform texture among patterns of a series and toward fine, rather than coarse, textures. Clearly, a broad and evolving literature exists on the use of areal symbols in cartography. How this body of research fits into the design of CRT area symbols is not so clear. Several problems are immediately apparent in applying this body of knowledge to dithered patterns on CRT monitors. First, the CRT works with emitted light, not reflected light as in the printed products used in the studies mentioned. The resolution of the CRT is fixed (on the TIPC there are approximately 75 pixels per inch horizontally and 43 pixels per inch vertically) while in printing, screens with different numbers of lines per inch are available. In addition, color is very available on the CRT and any number of versions of a map can be displayed virtually without increasing cost; great expense can be incurred when changing colors on printed maps. Also, with the default color of the CRT screen being black (it can be any other available color if the user specifies, however), dark is no longer "more", a tenet of cartographic convention associated with printed maps. DeSpite the fact that many accepted conventions and practices can not be directly translated to CRT images, previous research was helpful in designing the dithered patterns tested in this study. As Castner and Robinson (1969, pp. 10-11) suggest, the type of area symbols chosen for basic research “should be the simplest, most fundamental form of mark I arranged in the simplest, most uniform manner; only in this way can one eliminate, or reduce as much as possible, any distractions or undesirable 18 associations produced either by the character of the marks or their arrangement." Similarly, Jenks and Knos advocate use of regularly spaced dot patterns with fine texture. The patterns of this study were limited to dot patterns with the elements arranged in a triangular lattice (as shown in Figure 1). By using different densities of pixel spacing in the lattice, the patterns were also restricted to vary only in value. The dark colors (royal blue, red, and magenta) were dithered with white pixels; the light colors (white, green, yellow, and cyan) were dithered with black pixels. There is no reason that all of the eight colors on the monitor could not be combined in multiples or pairs, but the progression would not go from light to dark (and neutral to saturated or vice versa). Table 1 shows the horizontal and vertical increments of foreground pixels of patterns first chosen for possible inclusion in this study. These patterns were displayed on the CRT using a program similar to that in Appendix A and were judged (by me only) for suitability for mapping. Obvious problems with some of these patterns were immediately ap- parent. As Tufte (1983) forewarns, computer displays often contain "unintentional optical art" in the form of vibrating or moire patterns. These disturbing effects are one type of graphic noise that Tufte calls “chartjunk.” Slocum and McMaster (1986) have noted these effects also and state that “apparent pattern in a symbol implies a qualitative character- istic that (is) inappropriate for quantitative data.“ Several of the dithered patterns observed on the CRT monitor contained patterns and moirés, some more severe than others. In addition, the regular spacing of the pixels of the patterns eliminated all but one pattern between the values of 25% and 75%. TABLE 1 Distribution of Foreground and Background Pixels in Preliminary Dithered Patterns Light Foreground/Dark Background Pattern Number Horizontal Increment Vertical Increment Value % 1 (all pixels dark) 0.0 % 2 10 3 3.3 % 3 10 2 5.0 % 4 10 1 10.0 % 5 8 3 4.2 % 6 8 2 6.2 % 7 8 1 12.5 % 8 6 3 5.5 % 9 6 2 8.3 % 10 6 1 16.6 % 11 4 3 8.3 % 12 4 2 12.5 % 13 4 1 25.0 % 14 2 3 16.6 % 15 2 2 25.0 % 16 2 1 50.0 % Dark Foreground/Light Background 17 2 75.0 % 18 2 3 83.4 % 19 4 1 75.0 % 20 4 2 87.5 % 21 4 3 91.7 % 22 6 1 83.4 % 23 6 2 91.7 % 24 6 3 94.5 % 25 8 1 87.5 % 26 8 2 93.8 % 27 8 3 95.8 % 28 10 1 90.0 % 29 10 2 95.0 % 30 10 3 96.7 % 31 (all pixels light) 100.0 % 19 20 Figure 2 shows simulated examples of dithered patterns that required further consideration. The first pattern (4) is representative of a group (Patterns 4, 6, 25, and 28 from Table 1) that contained visible vertical stripes. The second (14) is representative of a group (Patterns 11, 14, 15, 17, 18 and 21) that had horizontal stripes. Patterns 13 and 19 had a diamond-shaped pattern within them. Numbers 15 and 17 had an "op art“ look and pattern 16 had elliptical shapes within it. The diamond patterns of 13 and 19 were judged to be acceptable, but the other patterns were dropped from further consideration. The elimination of the disturbing patterns from the study removed the 50% value pattern and left a gap in the value spectrum. All patterns between 25.0% and 75.0% value had been dropped. This problem was also encountered in the research of Plumb and Slocum (1986) on regularly spaced dot patterns for dot-matrix printers. To fill the gap in values, they designed new patterns. These patterns continued to have regularly spaced elements, but contained much more texture than the original patterns. The need for middle-range value patterns led me to design new pat- terns also. An alternative pattern was designed to replace the 50.0% pattern. New patterns were created to represent 40.0% and 60.0% values. All three of these patterns are simulated in Figure 3. Although these patterns do contain obvious textural differences from the other patterns, the new 50% value pattern seemed less disturbing than the pattern that it replaced. The 40% and 60% value patterns were designed to fill in the gaps in the value spectrum. Figure 4 shows simulations of the 23 patterns chosen to begin the study. Since the overriding questions were how many patterns would be dis- tinguishable and how successful would they be in a map context, 23 seemed a a) Pattern 4 0) Pattern 13 d) Pattern 15 Figure 2. Simulated dithered patterns with disturbing optical effects. a) Vertical stripes; b) Horizontal stripes; c) Diamond shapes; d) Op art; and e) Elliptical shapes. The figure was constructed with an Apple Macintosh computer and LaserWriter printer. which make the patterns look different than on the Texaslnstruments Professional Computer monitor. In particular. the optical effects of patterns 13, 15. and 16 were much more pronounced on the monitor. 21 5353i" III5$:IE:E$I :55; 1': I IIIII'I'IIIII'I I. I'.:"III+ " '1’?“ ii 1': .:- iii:- fiver." 'III'IIIIII I C) 60% Pattern Figure 3. Redesigned patterns with middle-range values (simulated). 22 4 (5%) 8 (12.5%) 16 (87 5%) 20 (95%) .. 12 (50%) ' 3 (4.2%) ' 11 (40% ' 19 (94.5%) 23 (100%) 23 14 (75%) 18 (93.8%) 22 (96.7%) 13 (60%) 17 (917%) 21 (95.9%) Figure 4. Test patterns for Study 1 (simulated). The numbers in parentheses indicate the value percentage of the patterns. 24 reasonably generous but not overwhelming number with which to work. The patterns range from 0% to 100% with most of the patterns under 25% or over 75%. The pixel arrangements of these patterns is shown in Table 2. The Slides Photographic slides were chosen as the medium for presenting the dithered patterns to human subjects. Slides have the advantages of being inexpensive, simple to make, and time-efficient to use for testing. In addition, they are a common medium for displaying the results of CRT mapping to large numbers of viewers. The way in which CRT monitors display an image determines how photo- graphs of the screen can be made. The CRT image is produced by an electron beam that quickly scans the surface of the tube. The beam produces only half of the image at a time, however. The beam scans every other line and then returns to scan the remaining lines. The appearance of a static image is created by the rapidity of these beam passes relative to the speed with which the eye-brain system processes the image. The interleaving process (i.e., display of the full image) takes between 1/25 and 1/30 second. If shutter speeds faster than 1/30 second are used to take the photograph, the image may be incomplete on the slide, but if slower speeds are used, at least one complete scan will occur while the shutter is open and a clear and complete image will result. An article in a photographic magazine (Pbgtg, Vol. 5, Issue 67) makes several suggestions for making successful slides from television screens with equipment most photographers possess. The article suggests darkening the room, eliminating light from windows and lamps to lessen reflections on the screen surface, and providing a black background to TABLE 2 Distribution of Foreground and Background Pixels in Dithered Patterns Chosen for Study Light Foreground/Dark Background Pattern Number Horizontal Increment Vertical Increment Value % 1 (all pixels dark) 0.0 % 2 10 2 3.3 % 3 8 3 4.2 % 4 10 2 5.0 % 5 6 3 5.5 % 6 8 2 6.2 % 7 6 2 8.3 % 8 4 2 12.5 % 9 6 1 16.6 % 10 4 1 25.0 % 11* 40.0 % 12** 50.0 % Dark Foreground/Light Background 13* 60.0 % 14 4 1 75.0 % 15 6 1 83.4 % 16 4 2 87.5 % 17 6 2 91.7 % 18 8 2 93.8 % 19 6 3 94.5 % 20 10 2 95.0 % 21 8 3 95.8 % 22 10 3 96.7 % 23 (all pixels light) 100.0 % * Foreground pixels distributed differently over 2 consecutive rows. Row 1 - foreground pixel every fifth pixel; Row 2 - two adjacent pixels of foregound color, followed by three adjacent pixels of background color. **A four-row repeat checkerboard pattern. Rows 1 and 2 - one pixel of foreground, one pixel of background; Rows 3 and 4 - one pixel of back- ground, one pixel of foreground. 25 26 flatten out the rounded shape of the screen. The effect of the shape of the screen can be also lessened by using a telephoto lens. It is necessary to use a tripod because of the slow shutter speed required. The camera back should be aligned parallel to the surface of the screen. With stationary objects, 1/8 second exposures are recommended. Bracketing exposure times using varying f—stop intervals ensures at least one good slide; F/5.6 is offered as a good starting point. For the test slides, I used Kodak Ektachrome (ASA 64) film, which produced slides with colors very similar to the monitor. This was con- firmed by Carter (1984, p. 13), who suggests that "daylight film has the proper balance for raster-scan CRTs." Using a camera with a 50mm lens and tripod, a complete set of slides was photographed at F/5.6 at 1 second for use in the study described in the next chapter. CHAPTER 3 EVALUATION, PART I: VALUE ESTIMATION, PATTERN DIFFERENTIATION, PREFERENCE The selection of dithered patterns for cartographic area symbols in this study, although based on the results of previous cartographic re- search has been intuitive. Many questions remain regarding the choice of appr0priate dithered patterns for cartographic use. The research discussed in this chapter investigates some perceptual properties of the chosen dithered patterns. In particular, value estima- tion of a set of monochromatic patterns in several color schemes, differ- entiability of the patterns from one another, and the subject's preference for particular color schemes used in the patterns will be studied in this chapter. Lem: TEST SLIDES Three types of slides containing images of dithered patterns display- ed on a CRT monitor were shown to subjects to solicit value estimation of the patterns and evaluate pattern differentiation. For convenience, these slide types will be called Type A, B, and C. Samples of the slide types are contained in an envelope on the back inside cover of the thesis. The Type A slide contained a set of seven royal blue/white dithered patterns ranging in value from 0% blue to 100% blue. Patterns 1 (0%), 2 (3.3%), 7 (8.3%), 12 (50%), 17 (91.7%), 22 (96.7%), and 23 (100%) were 27 28 displayed. The slide was used to assess subjects' perception of value percentage of the patterns presented. There was only one slide of Type A. Type 8 slides contained the same seven royal blue/white patterns as on the Type A slide below another set of identically formed patterns in another color scheme. These slides were used to determine if the value of the patterns was perceived differently using different colors in compar- ison to the royal blue/white patterns and if the new color scheme yielded more, less, or the same amount of differentiability between patterns. There were six slides of Type B. The color schemes of the other (non-royal blue/white) value scale were: 1) magenta/white; 2) yellow/black; 3) white/black; 4) cyan/black; 5) red/white; and 6) green/black. Type C slides contained two boxes filled with royal blue/white dithered patterns, which were labelled "A” and "B". Each pattern in the test was paired with itself and with each of the five most adjacent value percentage patterns. Table 3 shows the 123 pattern pairs made into slides. These slides were used to judge differentiability between the patterns chosen for the study. SUBJECTS The 64 subjects taking the test were undergraduate students, gradu- ate students, and faculty members of Michigan State University. Most subjects were affiliated with the Geography Department, and of these, many were cartography students. Thirty-six subjects were male; twenty- eight were female. Because of the large number of Type C slides, the slides were split into two groups for testing to reduce the response burden per subject. One group TABLE 3 Pattern Pairs Chosen for Study PATTERN NUMBER 16 17 18 19 20 21 22 23 10 11 12 13 14 15 9 8 X HNMQ‘LOQINGDOS O .—t 2 \O 30 of subjects was tested using the Type A slide, the six Type 8 slides, and 60 Type C slides. The other group was tested using the Type A slide, and 63 Type C slides. TESTING ENVIRONMENT Testing was conducted in the late afternoon in a small room darken- ed for viewing the slides. The slide projector was positioned approx- imately 8 feet away from the screen, which in combination with the focal length of the projector lens created an image approximately 24 by 16 inches on the screen. Subjects sat between 8 and 16 feet away from the screen. They were tested in small groups; no group contained more than 6 subjects. Testing lasted approximately 25 minutes. PROCEDURE After all subjects for a session were seated in the testing room, the door was closed and a brief explanation of the study was given. The subjects were told that this was an experiment about the use of patterns on CRT monitors and that they would look at slides of the patterns and answer some simple questions about them. Each subject was given a con- sent form to sign (See Appendix C), a test booklet (See Appendix D), and a pencil. They were told that they could ask questions at any time dur- ing the session. The room was darkened and the slide projector turned on. The sub- jects were asked if they felt that they could read the text on the test booklet and yet view the slides easily. Subjects were also told to write in the margins of the test booklet any comments they had on the testing itself, any vision problems they had (especially color deficiencies), and any preferences they had for 31 particular patterns or color schemes as they looked at the slides. The subjects were then instructed to read the first set of directions on the test booklet, and the first slide (Type A) was presented. Using this slide, subjects were to estimate the value percentages of the five patterns between solid color and solid white. After all subjects had completed their answers on the first slide, they were asked to read the second set of directions. (Note: The second group of subjects was not tested on this section.) As before, the sub- jects were to estimate the value percentages of five patterns. However, they were also asked to determine if the patterns in the new color scheme were more or less differentiable than the royal blue/white patterns, or about the same. Six slides were presented. Both testing groups were asked to read the next set of directions on the testing booklet. Subjects viewed slides with two dithered patterns to determine if the patterns contained the same value percentage of blue. If the patterns were considered to contain different percentages, subjects were to determine which of the patterns contained the higher percentage of blue. Subjects were verbally instructed to make a “first impression“ type of answer, and that each slide would be presented for only 10 seconds. Sixty Type C slides were shown to the first group of subjects; sixty-three to the second group. Results The results of the value estimation experiment of the royal blue/ white patterns are shown in Table 4. Using a two-tailed t-test with 60 degrees of freedom, all of the estimated means of value percentage were determined to be significantly different from the actual value at the TABLE 4 Estimates of Value Percentages of Royal Blue/White Patterns (N = 63) Pattern 2 Pattern 7 Pattern 12 Pattern 17 Pattern 22 (3.3%) (8.3%) (50.0%) (91.7%) (96.7%) '7 17.10 30.59 55.35 81.75 89.57 s 9.65 10.56 9.74 10.84 5.46 t -11.28* -16.75* -4.36* 7.29* 10.37* * p §_.025 TABLE 5 Estimates of Value Percentage for Different Colored Patterns (N = 34) Pattern Statistic Pat. 2 Pat. 7 Pat. 12 Pat. 17 Pat 22 Colors (3.3%) ,(8.3%) (50.0%) (91.7%) (96.7%) Magenta/ '7 17.88 28.97 58.82 84.82 90.21 white s 10.62 13.70 14.83 9.97 6.03 t -8001* -8080* ‘3047* 4002* 6028* Yellow/ 7' 15.88 30.29 58.53 82.22 89.00 black 5 10.48 14.92 16.07 12.56 7.16 t -7.00* -8.59* -3.09* 4.40* 6.27* White/ 7' 15.97 32.26 57.18 83.25 89.53 black 5 11.06 18.01 15.40 12.24 6.68 t -6.68* -7.76* -2.72* 3.85* 6.26* Cyan/ '7 15.09 28.12 54.71 81.74 89.00 black 5 11.91 14.53 11.07 13.34 7.44 t -5.77* -7.95* -2.48* 4.35* 6.03* Red/ 7' 14.97 28.38 55.35 83.15 89.68 white 5 10.11 13.18 13.90 11.57 5.99 t '6073* -8088* '2024* 5032* 6083* Green/ 7' 12.12 24.47 53.53 82.94 88.91 black s 10.22 13.79 15.50 12.89 7.65 t -5003* -6084* -1033 3096* 5094* * p §_.025 32 100% .1 90%;) Y 80% «)- 70% 1 I WA .1- 500/0 "' 400/0.- 30%" 20% ~- TOo/o "' J l l l A I 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Figure 5. Relationship between estimated and actual value percentage of royal blue/white patternsi 33 34 .025 level of significance. The graph in Figure 5 shows the relationship between the estimated and actual values of the patterns. Similar results were found for value estimation of identical patterns using different color schemes. The results are shown in Table 5. Only one estimated mean was not statistically different from its actual value percentage at the .025 level with 30 degrees of freedom. This was the 50% pattern on the green/black slide. All other estimations were signif- icantly different from the actual value. The normal approximation to the binomial distribution was used to test whether different color schemes were more, less, or about the same in differentiability compared to the royal blue/white patterns. The following formula was used: Z = (X - NP)/J NP (I-P) where X = the number of correct responses; N = the number of subjects; and P = the probability of a correct response occuring by chance. The probability of any answer (more, less, or same amount) is .33. Given an N of 34, at least 16 responses (Z = 1.75 [which is the first value ex- ceeding 1.65], p < .05) are required for statistical significance. The results are summarized in Table 6. As Table 6 shows, the patterns using the magenta/white, cyan/black, and green/black color schemes were found to be more differentiable than the royal blue/white scheme. The green/black scheme was perceived as being the most differentiable. The yellow/black and red/white color schemes appeared less differentiable than the royal blue/white scheme. TABLE 6 Perceived Amount of Differentiability Between Royal Blue/White Patterns and Those of Different Color Schemes Pattern More Less Same Colors Differentiable Differentiable Amount Magenta/ 19* 0 15 White Yellow/ 13 19* 2 Black White/ 10 11 13 Black Cyan/ 20* 8 6 Black Red/ 4 17* 13 White Green/ 25* 5 4 Black * P §_.05 N - 34 35 36 The normal approximation to the binomial distribution was also used to evaluate the 123 pattern pairs selected for investigation of pattern differentiability. However, the probability of determining whether a pattern is the same or different from another is .50. Given an N of 34 (the first group of subjects), at least 23 (Z = 2.05) correct responses are needed for statistical significance at the .05 level. Given an N of 30 (the second group of subjects), at least 23 (Z = 2.19) “I correct responses are needed for the same conclusion. Table 7 presents the number of correct responses for the 123 slides. Table 8 shows which rm; I pattern pairs were determined to be the same or different. 0f the patterns determined to be different from one another, all ordinal rankings of the value percentages of the pair were in the correct direction. That is, the pattern with the lowest value percentage of blue was perceived to be "bluer." Discussion The results of the value percentage estimates for the patterns show that subjects do not judge value of the royal blue/white patterns chosen for this study as a linear relationship. Similar to the results of Williams (1958), the values of patterns that were predominantly blue were overestimated; the values of predominantly white patterns were under- estimated. (Note that the measurement of the pattern itself was percent colored pixels, not a physical measurement of the amount of light.) Similar results were found for the patterns using other colors. Comparing all seven color schemes, the closest to actual value estimates were made on the green/black colored patterns. However, the presence of the original royal blue/white value scale on the Type 8 slides probably A.muwrm m_:p op ecoqmme uo: u_u pomnazm meow mm u z + em n z 9* om u z s wawN MN kmarsfim NN «NN#«OH *NN HN «ma «Ha «OH «wN CN *«oNkerrsmarewH««CN ma «NN¥«n~ +ON «m «35H rcN ma *HNwraN «Oasrwa «NH «we NH rmH ksakrNNrer «HH«¥ON ma rrwN ¥HN #NN *HN «Hg wsN mH ream rrisomerm*9Hm «eN «H **mm mwNwr¢mwer #wH sz mH «mNrrmm «omwwem «om «xN NH mmmzzz «omrrmm «omwrmm¥«¢m «RN HH zmmkk65 66. 665 56 56665656656 562 5 .6665 66: mmcoamme we“ sows; soc» 665 6:5 :6 66666—6 yo amass: use mmpocmu mmmmzpcmema cw conga: we» “muoz 66.N N6.N 56.6 6N.6 66.6 56.6 66.6 m 6.66 6N.N6 N.Ne No.66 66.5N 6.6m 6.6N N6 66.6N 66.6N 66.6N 6N.6N 66.66 66.66 66.66 .w 666 6N6 N66 666 NNN NNN meN 5656* 56 mm an 56 N6 66 66 5656 N5 Nm N6 46 em 56 6N 66 5N56 66 mm em 66 mm mm mm H655 55 6N 6N om mm 5N 65 ON ANHV NM 66 N 5m 56 Nm Nm 6 6 65 N6 56 N6 56 6N 6N 5N 5N5V mm em Nm em mm 66 56 5656 mm mm em 66 65 em 66 56.6 6 mm em 66 56 NN 66 Nm 56H6 56 em 66 N6 36 em 66 56 6 am .66 hm hm 66 mm 6m, . 6 6 65 6H, 6N 6N Hm, em 56 .mivtr 6N 65 6N 6N mm 66 mm 56-6 mm 36 NN mm mm 66 Nm .H6.6 6 565 +65 65 65 6N NN NN Aoev 6N 65 NN 6N 6N mm 66 56 6 56, mm em hm 66 am an 56.6 6 +65 N5 .65 +65 65 .65 NN 5N56 565 .65 565 6N5 5N 6N NN 565v 6 6N 6N 6N 5N 6N 6N em 5N5v .mm .66 Nm mm N6 .66 .em 56.6 m 56 cm 56 on mm em em 5N56 mm, mm, Nm Nm. 6m em .em .656.N Nm cm om New .lem, .66 dew .655 H. 56656\66N6 .56\=6666 .56\36556> .56\65563 65.63\.66z 6556=N661 .6z\6656.e 6265568 msmzum eopou Lea mmmcoqmwm pomceoo :66: NH NAm334 THEN GOTO 790 IF (I/HORIZONTALI) - INT(I/HORIZONTAL1)=O THEN PSET(I+ (HORIZONTALI/Z),J),FOREGROUNDI GOTO 790 IF I/HORIZONTALI - INT(I/HORIZONTAL1)=O THEN PSET(I,J), FOREGROUNDI NEXT I NEXT J RETURN REM Subroutine to label the boxes LOCATE 15,32 910 920 930 940 950 960 970 980 990 1000 1010 1020 1030 1040 1050 1060 1070 1080 1090 1100 1110 1120 1130 1140 1150 1160 62 PRINT “a b" RETURN REM Subroutine to plot boxes FOR I=1 TO 3 )) -(X(I+1).Y(I+1)).0 ELINE (X(I), Y(I LINE (X(4). Y(4 )) - (X(I),Y(1)).0 ( FOR I=5 T07 LINE (X(I), Y I)) - (X(I+1),Y(I+I)),O NEXT I LINE (X(8),Y(8)) ‘ (X(5)9Y(5))90 RETURN REM Subroutine to fill box 2 with dithered pattern INCR=0 FOR J=100 T0 160 IF J/VERTICALZ - INT(J/VERTICALZ) = 0 THEN 1150 INCR=INCR+1 FDR I=366 T0 474 IF INCR/2 - INT(INCR/Z) = 0 THEN GOTO 1130 IF I+HORIZONTAL2>474 THEN GOTO 1140 IF (I/HORIZONTAL2)-INT(I/HORIZONTAL2)=0 THEN PSET(I+ (HORIZONTAL2/2),J),FOREGROUNDZ GOTO 1140 IF I/HORIZDNTAL2 - INT(I/HORIZDNTALZ) = 0 THEN PSET(I,J), FOREGROUNDZ NEXT I NEXT J RETURN 10 20 30 40 50 60 7O 8O 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 260 270 280 290 300 310 320 330 340 350 360 370 380 390 400 410 420 APPENDIX B BASIC Computer Program to Create a Choropleth Map with Dithered Patterns CLS KOUNT=O DIM BACKGROUND(12),FOREGROUND(12),HSPACE(12),VSPACE(12) FOR I=1 T0 12 READ BACKGROUND(I),FOREGROUND(I),HSPACE(I),VSPACE(I) NEXT 1 DATA 1,1,10,10 DATA 1,7,10,2 DATA 1,7,6,2 DATA 1,7,4,2 DATA 1,7,6,1 DATA 1,7,4,1 DATA 1,7,40,40 DATA 1,7,50,50 DATA 7,1,40,40 DATA 7,1,6,I DATA 7,1,10,2 DATA 7,1,10,10 SE: Read the number of classes for the map REM Draw legend boxes and fill with appr0priate patterns FOR I=0 TO N LINE(135,106+I*12) - (165,106+(I-1)*12),3 NEXT I LINE(135,106) - (135,106+(I-1)*12),3 LINE(165,106) - (165,106+(I-1)*12),3 XMIN=135 XMAX=165 FOR Q=1 TO N LOCATE 9+Q,10 PRINT USING “ YMIN=106+(Q-l)*12 YMAX=106+Q*12 IF Q=1 THEN PATTERN=1 IF Q=2 THEN PATTERN=5 IF Q=3 THEN PATTERN=9 IF Q=4 THEN PATTERN=12 IF PATTERN=7 OR PATTERN=9 OR PATTERN=8 THEN GOTO 390 ELSE GOTO 410 IF PATTERN=8 THEN GOSUB 7000 ELSE GOSUB 6000 GOTO 420 GOSUB 4000 NEXT 0 class ##“,Q 63 430 440 450 460 470 480 490 500 510 520 530 540 550 560 570 589 590 600 610 620 630 640 650 660 670 680 690 700 710 720 730 740 750 760 770 780 790 800 810 820 830 840 850 860 870 880 890 64 FOR I=0 TO N LINE(135,106+I*12) - (165,106+I*12),O NEXT I LINE(135,106) - (135,106+(I-1)*12),O LINE(165,106) - (165,106+(I-1)*12),O DIM X(319),Y(319) DIM POINTDICT(SO) OPEN “B:MIPTS.DAT“ FOR INPUT AS #1 REM Read in the data from an external file FOR I=1 TO 319 INPUT #1, X(I),Y(I) NEXT I REM Variable COLOUR is the color of the state outline COLOUR=3 REM Now plot the state outline RESTORE 8000 FOR K=1 T0 85 READ NUMPOINTS FOR I=1 T0 NUMPOINTS READ POINTDICT(I) NEXT I FOR I=1 TO NUMPOINTS-1 LINE(X(POINTDICT(I)),Y(POINTDICT(I))) - (X(POINTDICT(I+1)), Y(POINTDICT(I+I))),COLOUR NEXT I LINE(X(POINTDICT(NUMPOINTS)),Y(POINTDICT(NUMPOINTS))) - (X (POINTDICT(I)),Y(POINTDICT(1))),COLOUR READ XMIN,XMAX,YMIN,YMAX,PATTERN IF PATTERN=7 OR PATTERN=9 OR PATTERN=8 THEN GOTO 700 ELSE GOTO 720 IF PATTERN=8 THEN GOSUB 7000 ELSE GOSUB 6000 GOTO 730 GOSUB 4000 NEXT K REM Redraw state outline in black RESTORE 8000 FOR K=1 T0 85 READ NUMPOINTS FOR I=1 TO NUMPOINTS READ POINTDICT(I) NEXT I FOR I=1 T0 NUMPOINTS-1 LINE(X(POINTDICT(I)),Y(POINTDICT(I))) - (X(POINTDICT(I+1)), Y(POINTDICT(I+1))),0 NEXT I LINEéX(POINTDICT(NUMPOINTS)),Y POINTDICT(NUMPOINTS))) - (X POINTDICT(I)),Y(POINTDICT 1))),O READ XMIN,XMAX,YMIN,YMAX,PATTERN NEXT K COLOR 0 LOCATE 1,1 END 2000 REM Pavlidis' algorithm for polygon fill 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100 2110 2120 2130 2140 2150 2160 2170 2180 2190 4000 4010 4020 4030 4040 4050 4060 4070 4080 4090 4100 4110 4120 4130 4140 4150 4160 4170 4200 4210 4220 4230 4240 4250 4260 4270 4280 4290 6000 6010 65 DY=-1 DX=1 ABOVE=O BELON=0 IF POINT(I-DX, J+DY)=3 THEN ABOVE=ABOVE+1 IF POINT(I-DX, J-DY)=3 THEN BELON=BELON+1 WHILE POINT(I,J)=3 IF POINT(I,J+DY)=3 AND POINT(I-DX,J+DY)<>3 THEN ABOVE=ABOVE+1 IF POINT(I,J-DY)=3 AND POINT(I-DX,J-DY)<>3 THEN BELON=BELON+1 I=I+1 KOUNT=KOUNT+1 IF PATTERN=7 AND KOUNT=5 THEN KOUNT=O IF PATTERN=8 AND KOUNT=4 THEN KOUNT=O IF PATTERN=9 AND KOUNT=5 THEN KOUNT=O GOTO 2070 HEND IF POINT(I-DX,J+DY)<>3 AND POINT(I,J+DY)=3 THEN ABOVE=ABOVE+1 IF POINT(I-DX,J-DY)<>3 AND POINT(I,J-DY)=3 THEN BELON=BELOH+1 RETURN REM 0 Subroutine to fill most polygons INC= FDR J=YMIN TO YMAX IF J/VSPACE(PATTERN)-INT(J/VSPACE(PATTERN))<>0 THEN INC=INC+1 IF VSPACE(PATTERN)=1 THEN INC=INC+1 COUNT=O I=XMIN WHILE I<=XMAX IF POINT(I,J)=3 GOTO 5020 IF COUNT/2-INT(COUNT/2)>O THEN GOSUB 4200 I=I+1 GOTO 4070 GOSUB 2000 IF ABOVE=1 AND BELOH=1 THEN COUNT=COUNT+1 GOTO 4070 NEND NEXT J RETURN REM Subroutine to fill patterns IF VSPACE(PATTERN)=1 GOTO 4230 IF J/VSPACE(PATTERN)-INT(J/VSPACE(PATTERN)=0 THEN 4280 IF INC/2-INT(INC/2)=0 THEN 4260 IF I/HSPACE(PATTERN)-INT(I/HSPACE(PATTERN))=0 THEN PSET(I,J), FOREGROUND(PATTERN) ELSE PSET(I,J),BACKGROUND(PATTERN) GOTO 4290 IF I/HSPACE(PATTERN)-INT(I/HSPACE(PATTERN))=.5 THEN PSET(I,J), FOREGROUND(PATTERN) ELSE PSET(I,J),BACKGROUND(PATTERN) GOTO 4290 PSET(I,J), BACKGROUND(PATTERN) RETURN REM Subroutine to fill 40% and 60% patterns INCR=O 6020 6030 6040 6050 6060 6070 6080 6090 6100 6110 6120 6130 6140 6150 6160 6170 6180 6190 6200 6210 6220 6230 6240 6250 7000 7010 7020 7030 7040 7050 7060 7070 7080 7090 7100 7110 7120 7130 7140 7150 7160 7170 7180 7190 7200 7300 7310 7320 7330 66 FOR J=YMIN TO YMAX COUNT=0 KOUNT=0 I=XMIN WHILE I<=XMAX IF POINT(I,J)=3 GOTO 6130 IF COUNT/2-INT(COUNT/2)>0 THEN GOSUB 6200 KOUNT=KOUNT+I IF KOUNT=5 THEN KOUNT=0 I=I+1 GOTO 6150 GOSUB 2000 IF ABOVE=1 AND BELOH=1 THEN COUNT=COUNT+1 NEND INCR=INCR+1 IF INCR=2 THEN INCR=0 NEXT J RETURN REM Subroutine to fill every fifth pixel IF INCR=0 THEN GOTO 6220 ELSE GOTO 6240 IF KOUNT=3 THEN PSET(I,J;,FOREGROUND(PATTERN) ELSE PSET(I,J), BACKGROUND(PATTERN GOTO 6250 IF KOUNT=O OR KOUNT=1 THEN PSET(I,J),FOREGROUND(PATTERN) ELSE PSET(I,J),BACKGROUND(PATTERN) RETURN REM Subroutine to fill 50% pattern INCR=0 FOR J=YMIN T0 YMAX COUNT=O I=XMIN KOUNT=O WHILE I<=XMAX IF POINT(I,J)=3 THEN 7130 IF COUNT/2-INT(COUNT/2)>O THEN GOSUB 7300 I=I+1 KOUNT=KOUNT+1 IF KOUNT=4 THEN KOUNT=O GOTO 7060 GOSUB 2000 IF ABOVE=I AND BELOH=1 THEN COUNT=COUNT+I GOTO 7060 NEND INCR=INCR+1 IF INCR=4 THEN INCR=0 NEXT J RETURN REM Subroutine to fill pattern IF INCR=0 OR INCR=1 THEN GOTO 7320 ELSE GOTO 7350 REM Fill blue, blue, white, white (for example) IF KOUNT=O OR KOUNT=1 THEN PSET(I,J),FOREGROUND(PATTERN) ELSE 67 PSET(I,J),BACKGROUND(PATTERN) 7340 GOTO 7370 7350 REM Fill white, white, blue, blue (for example) 7360 IF KOUNT=2 OR KOUNT=3 THEN PSET(I,J),FOREGROUND(PATTERN) ELSE PSET(I,J),BACKGROUND(PATTERN) 7370 RETURN 8000 DATA 5,311,141,142,143,310,532,567,130,144,4 8010 DATA 10,202,203,204,209,210,211,227,226,225,224,450,485,109,118,7 : (these lines contain the point dictionary) 8880 DATA 7,293,292,165,166,167,295,296,552,590,232,247,6 8890 DATA 5,313,317,221,241,229,424,452,145,157,9 PERCEPTION 0F DITHERED PATTERNS: Signed: APPENDIX C CONSENT FORM I have freely consented to take part in a scientific study being conducted by Ann Goulette under the supervision of Judy M. Olson, Professor. The study has been explained to me and I understand the explan- ation that has been given and what my participation will involve. I understand that I am free to discontinue my participation in the study at any time without penalty. I understand that the results of the study will be treated in strict confidence and that I will remain anonymous. Within these restrictions, results of the study will be made available to me at my request. I understand that my participation in the study does not guarantee any beneficial results to me. I understand that, at my request, I can receive additional explan- ation of the study after my participation is completed. Date: ADDENDA 1. 2. The purpose of this study is to improve the design of computer- generated maps. I understand that the expected length of my participation is 30 minutes. 68 APPENDIX 0 TEST BOOKLET AND INSTRUCTIONS Study 1 You will see a slide showing a set of boxes ranging in color from solid blue to solid white. If the left-most box is 0% blue and the right-most box is 100% blue, estimate the percentage of blue for the remaining boxes. 0% 100% Box 1 2 3 4 5 6 7 Now you will see sets of similar slides containing the same value scale you just saw and another scale using different colors. Please make estimations of the percentages for the new color schemes as you did for the first slide. Then, make a decision as to whether the patterns in the new color scheme are more or less differentiable or the same as the original set of patterns. 1. 0% 100% Less differentiable More differentiable Same amount of differentiation 6. 0% 100% Less differentiable More differentiable Same amount of differentiation 69 70 You will be shown slide of pairs of patterns. You should decide if the two patterns have the same percentage of blue or if they are distinctly different from one another. If they appear to be different, you should indicate which of the patterns (Pattern A or Pattern 8) seems to contain the greatest percentage of blue. Please make only one response per slide. 1. and B are the same is bluer than B is bluer than A and B are the same is bluer than B is bluer than A O CD>> WJ’) lll Ill 60. and B are the same is bluer than B is bluer than A W>> Study 2 INSTRUCTIONS You will be shown slides of a map of the State of Michigan. The per- son giving the test will point to a county on the map. You should match the pattern used to display that county with its appropriate legend box and write the number of the pattern on the test form. For example, if the county is colored green, find the green box in the legend and write the number of that box on this sheet. Please try to work quickly. SLIDE 1 SLIDE 34 1. Class 161. Class 2. Class 162. Class 163. Class SLIDE 2 164. Class 165. Class 3. Class 166. Class 4. Class "““‘ 5. Class SLIDE 35 6. Class . 167. Class 168. Class l l APPENDIX E DATA VALUES AND CLASSES s I S t a 3467954.8758567570878689894690970888429790 IpD. 1. 1.. .I. .I. 1. C3 am." .I. 14679548758567570878689894690970888429790 C 1.. .l. .l. .l. .l. S o S t a 34679548758567571888689894691971888429791 Ipp 1 1 1 1.. 1 ca mm.“ 0 S t s a5678657767667670777688785680870777528780 app. .I. 1.. .l. .I. 1.. .Ila mm." 8 “3456435545445457555466563467657555386567 O S t S 335806388686686828886008036020828883 20802 awp 1... 1.. 11.. 1. 11.1. .I. 1...]. 1.1. 1 ram." 6 “2345324434334346444355452356546444265456 O S t S 35592559959559592999592925522292999529992 app 1.. 1 .I. 1.. 1.1.1.. 1.. .I. .I. 1.3 mm." 4 02234223323223234333234342244434333243334 «mm. mu”9652794574765750454632419620250444922530 "1| thI e e S m xn r. a .13 n e e 0 N 093 d 0 e V In a 03 n n n VVaW nr 5 encatao 1|"Zk la aamca 68h“ 806 00 n CSIIPOdt m 1035 tnrgnIagVa IICO IprtfaentSWbTIShflaao esma noeeernar annhsnlyprnWtkO ede tqlgolnllcnbkak UCngterrVanraIsa8.13.131-Ctmnag oaIUPgnSOGCII OIIIInraaaeeraahthIr6.1a QIorrIounoorsaaa CAAAAAABBBBBBCCCCCCCCDDEEGGGGGHHHIIIIIJKK 71 72 % Change 4-class 6-class 8-class lO-class 12-class in Unem- map map map map map County Name loyment Cls. Pat. Cls. Pat. Cls. Pat. Cls. Pat. Cls. Pat. Kent 1.97 4' 12* 5’ 10’ 6* 8 8 9 9* T9 Keweenaw 0.5 4 12 6 12 7 10 9 11 10 10 Lake 8.8 2 5 2 3 3 5 3 4 4 4 Lapeer 6.7 2 5 3 6 4 6 5 6 6 6 Leelanau 0.8 4 12 6 12 7 10 9 11 10 10 Lenawee 7.6 2 5 3 6 4 6 4 5 5 5 Livingston 6.7 2 5 3 6 4 6 5 6 6 6 Luce 2.0 4 12 5 10 6 8 8 9 9 9 Mackinac 0.2 4 12 6 12 7 10 9 11 11 11 Macomb 6.7 2 5 3 6 4 6 5 6 6 6 Manistee 1.9 4 12 5 10 6 8 8 9 9 9 Marquette 7.5 2 5 3 6 4 6 4 5 5 5 Mason 6.9 2 5 3 6 4 6 5 6 6 6 Mecosta 1.7 4 12 5 10 6 8 8 9 9 9 Menominee 3.9 3 9 4 8 5 7 7 8 8 8 Midland 6.8 2 5 3 6 4 6 5 6 6 6 Missaukee 2.8 3 9 5 10 6 8 8 9 9 9 Monroe 6.6 2 5 3 6 4 6 5 6 6 6 Montcalm 7.9 2 5 3 6 4 6 4 5 5 5 Montmorency 17.0 1 1 1 1 1 1 1 1 I 1 Muskegon 4.9 3 9 4 8 5 7 6 7 7 7 Newaygo 8.2 2 5 3 6 3 5 4 5 5 5 Oakland 6.7 2 5 3 6 4 6 5 6 6 6 Oceana 4.9 3 9 4 8 5 7 6 7 7 7 Ogemaw -1.6 4 12 6 12 8 12 10 12 12 12 Ontonagon 6.1 2 5 3 6 4 6 5 6 6 6 Osceola 8.0 2 5 3 6 4 6 5 5 5 5 Oscoda 11.0 1 1 2 3 2 3 2 2 3 3 Ostego 0.3 4 12 6 12 7 10 9 11 11 11 Ottawa 1.9 4 12 5 10 6 8 8 9 9 9 Presque Isle 13.3 1 1 1 1 2 3 2 2 2 2 Roscommon 12.0 1 1 2 3 2 3 2 2 3 3 Saginaw 9.1 2 5 2 3 3 5 3 4 4 4 St. Clair 6.7 2 5 3 6 4 6 5 6 6 6 St. Joseph 6.7 2 5 3 6 4 6 5 6 6 6 Sanilac 9.8 2 5 2 3 3 5 3 4 4 4 Schoolcraft 3.6 3 9 4 8 5 7 7 8 8 8 Shiawasee 9.4 2 5 2 3 3 5 3 4 4 4 Tuscola 8.9 2 5 2 3 3 5 3 4 4 4 Washtenaw 2.9 3 9 5 10 6 8 8 9 9 9 Wayne 6.7 2 5 3 6 4 6 5 6 6 6 2 8 3 9 5 0 6 8 8 9 9 9 Wexford H SLIDE 1 (4-cl. Lake Washtenaw SLIDE 2 (6—cl. Gladwin Gratiot Ingham Oscoda SLIDE 3 (6-cl. Gratiot Ingham Oscoda Gladwin SLIDE 4 (8-cl. Alcona Cass Wexford Clinton SLIDE 5 (12-cl Ostego Shiawasee Presque Isle Roscommon Ottawa Gr. Traverse SLIDE 6 (6-cl. Oscoda Ingham Gladwin Gratiot SLIDE 7 (6-cl. Gratiot Gladwin Oscoda Ingham APPENDIX F SLIDE PRESENTATION ORDER IN STUDY 2 green) yellow) white) white) red) red) r. blue) SLIDE 8 (6-cl. cyan) Ingham Oscoda Gladwin Gratiot SLIDE 9 (IO-cl. green) Roscommon Genessee Ionia Ogemaw Montmorency Manistee Ostego Montcalm SLIDE 10 (8-cl. green) Cass Clinton Alcona Wexford SLIDE 11 (12-cl. cyan) Presque Isle Gr. Traverse Roscommon Ottawa Shiawasee Ostego SLIDE 12 (8-cl. yellow) Wexford Clinton Alcona Cass SLIDE 13 (8-cl. magenta) Alcona Clinton Cass Wexford 73 SLIDE 14 (4-cl. magenta) Lake Washtenaw SLIDE 15 (IO-cl. white) Genessee Ionia Manistee Montcalm Montmorency Ogemaw Ostego Roscommon SLIDE 16 (4-cl. yellow) Washtenaw Lake SLIDE 17 (12-cl. white) Gr. Traverse Ostego Ottawa Presque Isle Roscommon Shiawasee SLIDE 18 (8-cl. cyan) Wexford Clinton Cass Alcona SLIDE 19 (6-cl. green) Gratiot Oscoda Gladwin Ingham SLIDE 20 (12-cl. green) Presque Isle Shiawasee Ostego Roscommon Gr. Traverse Ottawa SLIDE 21 (12-cl. magenta) Ottawa Ostego Shiawasee Presque Isle Gr. Traverse Roscommon SLIDE 22 (IO-cl. red) Ostego Genessee Manistee Roscommon Montcalm Montmorency Ionia Ogemaw SLIDE 23 (12-cl. r.blue) Roscommon Presque Isle Ottawa Gr. Traverse Ostego Shiawasee SLIDE 24 (6-cl. magenta) Oscoda Gladwin Gratiot Ingham SLIDE 25 (IO-cl. cyan) Manistee Ostego Montmorency Genessee Ogemaw Roscommon Ionia Montcalm SLIDE 26 (4-cl. r. blue) Lake Washtenaw 74 SLIDE 27 (IO-cl. r. blue) SLIDE 34 (12-cl. yellow) Montcalm Ostego Montmorency Genessee Roscommon Manistee Ogemaw Ionia SLIDE 28 (8-cl. red) Wexford Alcona Cass Clinton SLIDE 29 (IO-cl. magenta) Manistee Roscommon Montcalm Ogemaw Montmorency Ostego Genessee Ionia SLIDE 30 (IO-cl. yellow) Roscommon Ostego Ogemaw Montmorency Montcalm Manistee Ionia Genessee SLIDE 31 (8-cl. r. blue) Clinton Alcona Wexford Cass SLIDE 32 (4-cl. cyan) Washtenaw Lake SLIDE 33 (4-cl. red) Lake Washtenaw Ostego Shiawasee Presque Isle Roscommon Ottawa Gr. Traverse SLIDE 35 (4-CT. white) Washtenaw Lake BIBLIOGRAPHY BIBLIOGRAPHY Brassel, K. A model for automatic hill-shading. The American Car- tographer, Vol. 1(1), 1974, pp. 15-27. Carter, J. R. Computer Mapping: Progress in the '80's. Washington, DC: Association of American Geographers, Resource Publications in Geography, 1984. Castner, H. W., & Robinson, A. H. Dot area symbols in cartography: The influence of pattern on their perception. Technical Mono- graph, No. CA-4. Washington, DC: American Congress of Surveying and Mapping, I969. Dent, B. 0. Principles of Thematic Map Design. Reading, MA: Addison- Wesley, 1985. Foote, K. E. Reviews of geographic software: Atlas AMP. The Profes- sional Geographer, Vol. 37(4), 1985, pp. 492-493. Gossette, F. Software review: PRODESIGN II: The easy to use CAD sys- tem. The American Cartographer, Vol. 13(2), 1986, pp. 168-171. Gr00p, R. E., & Smith, R. M. Matrix line printer maps. The American Cartographer, Vol. 9(1), 1982, pp. 19-24. Gr00p, R. E. Software review: Desktop Information Display System (DIDS) Version 2.0. The American Cartographer, Vol. 12(2), 1985, pp. 174-176. Haber, R. N., & Wilkinson, L. Perceptual components of computer dis- plays. IEEE Computer Graphics and Applications, Vol. 2(3), 1982, pp. 23‘350 Hobbs, L. C. Computer graphics display hardware. IEEE Computer Graph- ics and Applications, Vol. 1(1), 1981, pp. 25:39. Jenks, G. F., & Knos, D. S. The use of shaded patterns in a graded series. Annals of the American Association of Geographers, Vol. 51, 1961, pp.316-334. Jensen, J. R. Software review: MICROPIPS Personal Image Processing System/The Telesys Group, Inc. The American Cartographer, Vol. 13(1), 1986, pp. 81-85. McGranaghan, M. Symbol and background value effects in chor0pleth maps for CRT display. Technical Papers 45th Annual Meeting. ACSM, Washington, DC, 1985, pp. 3564363. 75 76 McGranaghan, M. Interaction of saturation and value in color choropleth maps for CRT display. Technical Papers, 46th Annual Meeting. ACSM, Washington, DC, 1986, pp. 76:84} Machover, C. CRT graphics terminals. Computer Graphics and Art, May, 1977, pp. 16-29. Michigan Statistical Abstract. Detroit, MI: Wayne State University *Press, 1981. Morrison, J. L., & Mendoza, G. Software review: AutoCAD Drafting Pack- age. The American Cartographer, Vol. 13(3), 1986, pp. 269-274. Murch, G. The effective use of color: Perceptual principles. Tek- nigues, Vol. 8(1), 1984, pp. 4-9. Pavlidis, T. Filling algorithms for raster graphics. Computer Graphics and Imagg Processing, Vol. 10, 1979, pp. 126-141. Pavlidis, T. Algorithms for Graphics and Image Processing. Rockville, MD: Computer SCience Press, 1982. Photo, Vol. 5, Issue 67, pp. 1841-1843. Plumb, G. A., & Slocum, T. A. Alternative designs for dot-matrix printer maps. The American Cartographer, Vol. 13(2), 1986, pp. 121-1330 Ryan, 0. Computer Programming for Gpaphical Dipplays. Monterey, CA: Brofls/CoTe Engineering DivisiFn, 1983. Slocum, T. A., & McMaster, R. B. Gray tone versus line plotter area symbols: A matching experiment. The American Cartographer, Vol. 13(2), 1986, pp. 151-164. Smith, R. M. Improved areal symbols for computer line-printed maps. The American Cartogpapher, Vol. 7(1), 1980, pp. 51-57. Taylor, D. R. F. Some conclusions and new challenges. In Taylor, D. R. F. (ed.), Graphic Communication and Design in Cartography. New York: John Wiley and’Sons, 1983, pp. 285-307. Truckenbrod, J. R. Effective use of color in computer graphics. Com- pgter Graphics, Vol. 15(3), 1981, pp. 83-90. Tufte, E. R. The Visual Display of Quantitative Information. Cheshire, CN: Graphics Press, 1983. White, 0. The ODYSSEY System for display and analysis of geographic information. In Moore, D. A. (ed.), Cartographic and Statistical Data Bases and MappinggSoftware. Cambridge, MA: Harvard Library of Computer GraphiCS, 1980, pp. 225-254. 77 Williams, R. L. Map symbols: Equal-appearing intervals for printed screens. Annals of the Association of American Geographers, V01. 48, 1958’ pp. IH-lggo .. “Wei“ ‘ / w; 10? g K. “@963 Ln