View - The Ohio State University

3 downloads 0 Views 1MB Size Report
School of The Ohio State University. By .... Columbus, Ohio, U.S.A. ...... Table 4.22: More frequent game players had fewer navigation errors than less frequent ...
INVESTIGATING WAYFINDING USING VIRTUAL ENVIRONMENTS

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University By Ebru Cubukcu, B.C. P. (Hons.), M.C.R.P. ***** The Ohio State University 2003

Approved by Dissertation Committee: Professor Jack L. Nasar, Adviser

Adviser

Professor Steven I. Gordon

Graduate Program in

Professor Kenneth Pearlman

City and Regional Planning Program

ABSTRACT

Wayfinding is the spatial knowledge about one’s current location, destination, and the spatial relation between them. Wayfinding problems threaten people’s sense of well-being, and cause loss of time and money. Designers and planners can improve wayfinding when they understand how physical environmental factors affect people’s wayfinding performance. This study explores the effect of personal and physical environmental characteristics on wayfinding performance. The personal characteristics include gender, age, and familiarity. The physical environmental characteristics include plan layout complexity, physical differentiation and its components vertical and horizontal differentiation. The experiment had eighteen (2 x 3 x 3) simulated environments, with two plan layouts (complex and simple), three kinds of vertical differentiation (no differentiation, object landmarks, and building landmarks) and three kinds of horizontal differentiation (no differentiation, road width variation, road pavement variation), and it also had four different question orders.

ii

166 volunteers (98 male, 68 female) were tested individually. Participants were randomly assigned to one of the question orders and to one of the simulated environments with the constraint that there would be equal number of people in survey types, in plan layout conditions, in vertical differentiation conditions, and in horizontal differentiation conditions. The experiment had a learning phase and a test phase. In the learning phase, participants actively explored one of the simulated environments at their leisure up to four minutes. In the test phase the participants completed three spatial knowledge tasks (a direction estimation task, a navigation task, and a sketching task) and a survey which had questions on gender, age, frequency of playing computer game, realism of the simulated environment judgement and wayfinding strategies used in the navigation task. As expected, the Simple layouts, Higher Physical Differentiation, Vertical or Horizontal differentiation yielded better wayfinding performance than Complex layouts, Lower Physical differentiation, and No Vertical or Horizontal differentiation. Males performed better than Females, and performance improved with Familiarity.

iii

DEDICATION

To my parents and my husband

iv

ACKNOWLEDGMENTS

I would like to express my deepest appreciation to my adviser, Professor Jack L. Nasar, for his expert and timely advice, guidance, inspirations, and encouragement throughout the research. I thank Professors Steven I. Gordon and Kenneth Pearlman for serving on the advisory committee and providing their helpful insights and comments. I would also like to acknowledge the assistance of Dr. Harry Heft for his helpful comments on the early draft of this dissertation. I thank Dr. Peter Hecht for allowing me to involve in wayfinding projects, which I truly enjoyed and benefited, towards the end of this dissertation. I would like to thank Dokuz Eylul University for the scholarship, and Selin Koroglu and Filiz Dincyigit for handling all the administrative work in Turkey regarding this scholarship. I also thank the Center for Mapping for the assistantship. I am also grateful to Jenny Klein, from the Office of Residence Life, who was very helpful in providing the site to conduct the surveys, and to people who participated as respondents in this study. I also extend my gratitude to Misun Hur,

v

In-Young Yeo, and my other fellow Doctoral students in the Department of City and Regional Planning for their positive reinforcement; to students in the Advanced Computing Center for the Arts and Design (ACCAD) for their help to discover the simulation tool in this dissertation; and to friends I met in Columbus for their friendship. Finally, I would like to thank my parents, Fulden and Ziya Demirayak, my brother, my sisters-in-law and my niece, Yasemin, for being everlasting supporters of my studies and believing in my capabilities. My special thanks go to my husband, Mert, for everything.

vi

VITA February 24, 1974.………………….Born – Antalya, Turkey 1997…………………………………B.C.P. (Hons.), City Planning School of Architecture Middle East Technical University Ankara, Turkey. 2001…………………………………M.C.R.P., City and Regional Planning Austin E. Knowlton School of Architecture The Ohio State University Columbus, Ohio, U.S.A. 1998-present………………………...Graduate Research Associate Dokuz Eylul University Izmir, Turkey. 2002-present...………………………Graduate Research Associate The Center For Mapping The Ohio State University Columbus, Ohio, U.S.A.

FIELDS OF STUDY Major Field: City and Regional Planning Minor Field: Environmental Psychology

vii

TABLE OF CONTENTS Page Abstract ......................................................................................................................ii Dedication ................................................................................................................. iv Acknowledgments ...................................................................................................... v Vita ...........................................................................................................................vii List of Tables............................................................................................................xii List of Figures .......................................................................................................xviii Chapters 1 Introduction ............................................................................................................. 1 2 Literature Review .................................................................................................... 5 2.1 The Concepts Related to Wayfinding ............................................................... 6 2.2 Significance of the Wayfinding Research......................................................... 9 2.3 The Physical Environmental and Personal Characteristics Affecting Wayfinding Behavior ...................................................................................... 10 2.3.1 The Plan Layout........................................................................................ 11 2.3.2 The Level of Physical Differentiation ...................................................... 12 2.3.3 The Vertical Differentiation ..................................................................... 13 2.3.4 The Horizontal Differentiation ................................................................. 15 2.3.5 Age............................................................................................................ 15 2.3.6 Gender ...................................................................................................... 18

viii

2.3.7 Familiarity (Experience)........................................................................... 21 2.3.8 Summary of Factors Effecting Wayfinding Performance ........................ 21 2.4 Tools to Simulate Environment ...................................................................... 22 2.4.1 Use of Photographs and Simulation Booth............................................... 23 2.4.2 Full Scale Models ..................................................................................... 25 2.4.3 Small Scale Models .................................................................................. 26 2.4.4 Computer Models (Virtual Environments)............................................... 28 2.4.5 Summary of Tools for Studying Wayfinding ........................................... 30 2.5 Measures of Wayfinding Performance ........................................................... 31 2.5.1 Self Report Tests ...................................................................................... 31 2.5.2 Memory Tests ........................................................................................... 32 2.5.3 Recognition Tests ..................................................................................... 33 2.5.4 Spatial Orientation Tests .......................................................................... 34 2.5.5 Navigation Tests ....................................................................................... 39 2.5.6 Summary of Wayfinding Measures.......................................................... 41 3 Methodology ......................................................................................................... 43 3.1 General Procedures and Equipment................................................................ 43 3.1.1 Introductory Procedures ........................................................................... 43 3.1.2 Equipment and Setting.............................................................................. 44 3.2 Virtual Environments...................................................................................... 44 3.2.1 Software.................................................................................................... 44 3.2.2 Physical Environmental Characteristics ................................................... 45 3.2.3 Realism of Virtual Environments Judgment ............................................ 57 3.3 Participants and Group Demographics ........................................................... 60 3.4 Experimental Procedures ................................................................................ 63 3.5 Measures ......................................................................................................... 66 3.5.1 Learning Phase ......................................................................................... 66 3.5.2 Test Phase ................................................................................................. 67 4 Results ................................................................................................................... 71 4.1 Relation Between Different Measures of Wayfinding Performance .............. 71 4.2 Relation Between the Self-Reported Navigation Strategy and Different Tasks Measuring Wayfinding Performance .................................................... 72 ix

4.3 The Effect of Physical Environmental and Personal Characteristics on Wayfinding Performance ................................................................................ 74 4.3.1 General Summary ..................................................................................... 75 4.3.2 Statistical Results...................................................................................... 86 4.3.2.1 Overall Spatial Awareness ................................................................. 88 4.3.2.2 Direction Estimation Task.................................................................. 94 4.3.2.3 Navigation Task ................................................................................. 99 4.3.2.4 Sketching Task ................................................................................. 104 5 Conclusion........................................................................................................... 110 Bibliography........................................................................................................... 121 Appendices: Appendix A The written description about the study ............................................ 136 Appendix B The effect of interaction between physical environmental factors on error scores ............................................................................................................. 137 Appendix B.1 General Linear Models on Comprehensive Measure (Overall Spatial Awareness).............................................................. 138 Appendix B.2 General Linear Models on Direction Estimation Task ....... 140 Appendix B.3 General Linear Models on Navigation Task....................... 142 Appendix B.4 General Linear Models on Sketching Task......................... 144 Appendix C The statistical analyses on specific measures (error scores) of navigation and sketching task ................................................................................ 146 Appendix C.1 The Specific Measures (Error Scores) of Navigation Task ................................................................................................... 147 Appendix C.2 The Specific Measures (Error Scores) of Sketching Task .. 151 Appendix D The success scores for various tasks.................................................. 156 Appendix D.1 The Measures of Success Scores ........................................ 157 Appendix D.2 The Analyses of Overall Spatial Awareness Success......... 160 Appendix D.3 The Analyses of Direction Success .................................... 162 Appendix D.4 The Analyses of Overall Navigation Success..................... 165 Appendix D.5. The Analyses of Overall Sketching Success ..................... 168

x

Appendix D.6. The Statistical Analyses on Specific Measures (Success Scores) of Navigation Task ............................................................... 171 Appendix D.7. The Statistical Analyses on Specific Measures (Success Scores) of Sketching Task................................................................. 175

xi

LIST OF TABLES Table

Page

Table 3.1: Level of differentiation was determined by the presence of vertical and horizontal differentiation ........................................................................... 48 Table 3.2: The means of judged realism across individual characteristics .............. 58 Table 3.3: The means of judged realism across physical environmental characteristics ................................................................................................... 59 Table 3.4: The distribution of participants across eighteen environments............... 62 Table 3.5: The distribution of participants across survey and environment conditions ......................................................................................................... 62 Table 4.1: Direction estimation, navigation and sketching scores had a statistically significant correlation with one another........................................ 72 Table 4.2: Remembering the correct turns strategy was associated with remembering the number of streets, buildings to pass strategy and keeping track of general directions strategy..................................................... 73 Table 4.3: Different tasks related to different navigation strategies. ....................... 73 Table 4.4: The four tests repeated for each task....................................................... 75 Table 4.5: The significance of plan layout effect on various tasks .......................... 77 Table 4.6: The significance of physical differentiation effect on various tasks....... 79 Table 4.7: The significance of vertical differentiation effect on various tasks ........ 79 Table 4.8: The significance of horizontal differentiation effect on various tasks .................................................................................................................. 81 Table 4.9: The significance of gender effect on various tasks ................................. 83 Table 4.10: The significance of age effect on various tasks .................................... 84 Table 4.11: The significance of familiarity effect on various tasks ......................... 84 Table 4.12: The significance of exploration speed effect on various tasks.............. 85

xii

Table 4.13: The significance of game playing effect on various tasks .................... 86 Table 4.14: General Linear Models on overall spatial awareness error for the first set of analyses. .......................................................................................... 88 Table 4.15: General Linear Models on overall spatial awareness error for the second set of analyses....................................................................................... 89 Table 4.16: As frequencies of game playing increased overall spatial awareness errors decreased. ............................................................................. 93 Table 4.17: General Linear Models on direction error for the first set of analyses. ........................................................................................................... 94 Table 4.18: General Linear Models on direction error for the second set of analyses. ........................................................................................................... 95 Table 4.19: As game playing increased direction errors decreased ......................... 98 Table 4.20: General Linear Models on overall navigation error for the first set of analyses. ..................................................................................................... 100 Table 4.21: General Linear Models on navigation error for the second set of analyses. ......................................................................................................... 101 Table 4.22: More frequent game players had fewer navigation errors than less frequent game players. ................................................................................... 103 Table 4.23: General Linear Models on overall sketching error for the first set of analyses. ..................................................................................................... 105 Table 4.24: General Linear Models on overall sketching error for the second set of analyses................................................................................................. 106 Table 4.25: More Frequent Game Players had fewer sketching errors than Less Frequent game players ........................................................................... 109 Table 5.1: The significance of the effects of physical environmental characteristics on various measures of wayfinding performance................... 114 Table 5. 2: The significance of the effects of personal characteristics on various measures of wayfinding performance................................................ 116 Table B.1: General Linear Models with the interaction between plan layout and level of physical differentiation............................................................... 138 Table B.2: General Linear Models with the interaction between plan layout and vertical and horizontal differentiation. .................................................... 139 Table B.3: General Linear Models with the interaction between plan layout and level of differentiation. ............................................................................ 140 Table B.4: General Linear Models with the interaction between plan layout and vertical and horizontal differentiation. .................................................... 141 xiii

Table B.5: General Linear Models with the interaction between plan layout and level of differentiation ............................................................................. 142 Table B.6: General Linear Models with the interaction between plan layout and vertical and horizontal differentiation ..................................................... 143 Table B.7: General Linear Models with the interaction between plan layout and level of differentiation ............................................................................. 144 Table B.8: General Linear Models with the interaction between plan layout and vertical and horizontal differentiation ..................................................... 145 Table C.1: General Linear Models on Speed for the first set of analyses.............. 147 Table C.2: General Linear Models on Speed for the second set of analyses. ........ 148 Table C.3 General Linear Models on Turn Error for the first set of analyses. ...... 148 Table C.4: General Linear Model on Turn Error for the second set of analyses. .. 149 Table C. 5: General Linear Model on Distance Error for the first set of analyses. ......................................................................................................... 149 Table C.6: General Linear Model on Distance Error for the second set of analyses. ......................................................................................................... 150 Table C.7: Binary Logistic Regression on MARKET Sign Location Error (at an intersection or on the road) for the first set of analyses............................. 151 Table C.8: Binary Logistic Regression on MARKET Sign Location Error (at an intersection or on the road) for the second set of analyses........................ 152 Table C.9: General Linear Model MARKET Sign Distance Error for the first set of analyses................................................................................................. 152 Table C.10: General Linear Model MARKET Sign Distance Error for the second set of analyses..................................................................................... 153 Table C.11: General Linear Model on Route Turn Error for the first set of analyses. ......................................................................................................... 153 Table C.12: General Linear Model on Route Turn Error for the second set of analyses. ......................................................................................................... 154 Table C.13: General Linear Model on Route Segment Error for the first set of analyses. ......................................................................................................... 154 Table C.14: General Linear Model on Route Segment Error for the second set of analyses. ..................................................................................................... 155 Table D.1: In simple environments more participants were successful in more tasks than complex ones. ................................................................................ 160

xiv

Table D.2: As differentiation increased from Low to Moderate to High, the percentage of respondents successfully completed more tasks increased...... 160 Table D.3: In environments with vertical differentiation more participants were successful in more tasks than those without it....................................... 160 Table D.4: In environments with horizontal differentiation more participants were successful in more tasks than those in environments without it............ 161 Table D.5: Males showed a higher success rate than females................................ 161 Table D.6: More frequent game players showed higher success rate than less frequent game players. ................................................................................... 161 Table D.7: In simple environments more participants were successful than in complex ones.................................................................................................. 162 Table D.8: As differentiation increased from Low to Moderate to High, the percentage of successful respondents increased............................................. 162 Table D.9: In environments with vertical differentiation more participants were successful than in environments without it............................................ 162 Table D.10: In environments with horizontal differentiation more participants were successful than in environments without it and road pavement variation produced better success rates. ......................................................... 162 Table D.11: Males showed a higher success rate than females.............................. 163 Table D.12: More frequent game players showed higher success rates than less frequent game players. ............................................................................ 163 Table D.13: Binary Logistic Regression on direction success for the first set of analyses. ......................................................................................................... 163 Table D.14: Binary Logistic Regression on direction success for the second set of analyses................................................................................................. 164 Table D.15: In Simple environments more participants were successful than in complex ones.................................................................................................. 165 Table D.16: As differentiation increased from Low to Moderate to High, the percentage of successful respondents increased............................................. 165 Table D.17: In environments with vertical differentiation more participants were successful than in environments without it............................................ 165 Table D.18: In environments with horizontal differentiation more participants were successful than in environments without it............................................ 165 Table D.19: Males showed a higher success rate than females.............................. 166 Table D.20: More frequent game players showed higher success rate than less frequent game players. ................................................................................... 166 xv

Table D.21: Binary Logistic Regression on overall navigation success for the first set of analyses. ........................................................................................ 166 Table D.22: Binary Logistic Regression on overall navigation success for the second set of analyses..................................................................................... 167 Table D.23: In Simple environments more participants were successful than in complex ones.................................................................................................. 168 Table D.24: As differentiation increased from Low to Moderate to High, the percentage of successful respondents increased............................................. 168 Table D.25: In environments with vertical differentiation more participants were successful than in environments without it............................................ 168 Table D.26: In environments with horizontal differentiation more participants were successful than in environments without it............................................ 168 Table D.27: Males showed a higher success rate than females.............................. 169 Table D.28: More frequent game players showed higher success rate than less frequent game players. ................................................................................... 169 Table D.29: Binary Logistic Regression on overall sketching success for the first set of analyses. ........................................................................................ 169 Table D.30: Binary Logistic Regression on overall sketching success for the second set of analyses..................................................................................... 170 Table D.31: Binary Logistic Regression on success based on speed for the first set of analyses. ........................................................................................ 171 Table D.32: Binary Logistic Regression on success based on speed for the second set of analyses..................................................................................... 172 Table D.33: Binary Logistic Regression on success based on turn error for the first set of analyses. ........................................................................................ 172 Table D.34: Binary Logistic Regression on success based on turn error for the second set of analyses..................................................................................... 173 Table D.35: Binary Logistic Regression on success based on distance error for the first set of analyses. .................................................................................. 173 Table D.36: Binary Logistic Regression on success based on distance error for the second set of analyses............................................................................... 174 Table D.37: Binary Logistic Regression on success based on map selection for the first set of analyses. .................................................................................. 175 Table D.38: Binary Logistic Regression on success based on map selection for the second set of analyses............................................................................... 176

xvi

Table D.39: Binary Logistic Regression on success based on locating MARKET sign exactly for the first set of analyses........................................ 177 Table D.40: Binary Logistic Regression on success based on locating MARKET sign exactly for the second set of analyses................................... 178 Table D.41: Binary Logistic Regression on success based on drawing the sequence of route turns for the first set of analyses. ...................................... 179 Table D.42: Binary Logistic Regression on success based on drawing the sequence of route turns for the second set of analyses................................... 180 Table D.43: Binary Logistic Regression on success based on drawing the route segments for the first set of analyses..................................................... 181 Table D.44: Binary Logistic Regression on success based on drawing the route segments for the second set of analyses................................................ 182

xvii

LIST OF FIGURES Figure

Page

Figure 2.1: The hierarchical structure of the study area, wayfinding……………... 4 Figure 2.2: Full scale models (Sanoff 1991)………………………………………. 22 Figure 2.3: The left one is the simulator at Lund (Sanoff 1991; p 146), right one is the simulator at the Institute of Urban and Regional Development at the University of California at Berkley. (Altman and Wohlwill, 1977; p.81)……………………………………………………………………….…… 23 Figure 3.1: The same house plan was repeated in all environments. From top left moving clockwise images show plan view (not seen by subjects), front view, right view and left view...……………………………………………………… 41 Figure 3.2: The arrows at intersections showed possible directions one can take and a message reminds users that they can change direction…………….…… 42 Figure 3.3: An example for calculating Interconnection density (ICD) value…….. 43 Figure 3.4: The plan layout and schematic drawings of the simple and complex settings in O’Neill’s study……………………………………………………... 44 Figure 3.5: The plan layout and schematic drawings of the simple and complex settings in this study……………………………………………………….…... 45 Figure 3.6: The START and MARKET signs in all environments………………... 46 Figure 3.7: Environments with vertical differentiation had two types of landmarks. TYPE A had four object landmarks shown from top left moving clockwise (one kind of lamp, another kind of lamp, a flower pot and a flag) at choice points. (All environments were in full color)…………………………. 47

xviii

Figure 3.8: Environments with vertical differentiation had two types of landmarks. TYPE B had four building landmarks that differ from one another and the surrounding buildings shown from top left moving clockwise (a gray brick building, an orange brick building, a white building and a yellow building) at choice points. (All environments were in full color)…………….. 48 Figure 3.9: The location of landmarks in the Simple and Complex environments... 49 Figure 3.10: In environments with road hierarchy the most efficient route between START and MARKET signs were wide or had asphalt pavement and all other roads were narrow or had cobblestone pavement……………………. 50 Figure 3.11: Environments with horizontal differentiation had two types of road hierarchy. For TYPE A (left column) road width varied and for TYPE B (right column) road pavement varied………………………………………….. 51 Figure 3.12: In the sketching tests, participants were asked to pick one of the four maps, that they thought best represents the environment they experienced. (Top row: A = Correct Complex Plan, B = Correct Simple Plan; bottom row: C = Distracter for Complex Plan, D = Distracter for Simple Plan.)…………... 59 Figure 4.1: The standardized mean error scores for each task is higher in the Complex environments than in the Simple ones………………………………. 77 Figure 4.2: The mean error scores for each task increases as level of physical differentiation decreases from High to Moderate to Low……………………... 78 Figure 4.3: The mean error scores for each task is lower in environments in which vertical differentiation is Present than the ones in which vertical differentiation is Absent……………………………………………………….. 80 Figure 4.4: The mean error scores for each task is lower in environments in which horizontal differentiation is Present than the ones in which horizontal differentiation is Absent……………………………………………………….. 81 Figure 4.5: Males had fewer mean errors than Females in all three tasks………… 82 Figure 4.6: The standardized mean error scores in each task decreases as game playing frequency increases…………………………………………………… 86 Figure 4.7: The significance of difference across different physical differentiation conditions on spatial awareness. Overlapping boxes indicate an insignificant difference and separate boxes indicate a significant difference……………….. 91

xix

Figure 4.8: The significance of difference across different Vertical Differentiation conditions on overall spatial awareness Overlapping boxes indicate an insignificant difference and separate boxes indicate a significant difference…. 92 Figure 4.9: The significance of difference across different Horizontal Differentiation conditions on overall spatial awareness. Overlapping boxes indicate an insignificant difference and separate boxes indicate a significant difference………………………………………………………………………. 92 Figure 4.10: The significance of difference between different types of horizontal differentiation conditions (Overlapping boxes indicate an insignificant difference and separate boxes indicate a significant difference) on direction estimation……………………………………………………………………… 98 Figure 4.11: The significance of difference between different types of Vertical differentiation conditions (Overlapping boxes indicate an insignificant difference and separate boxes indicate a significant difference) on direction estimation……………………………………………………………………… 109

xx

CHAPTER 1 INTRODUCTION

Wayfinding requires knowledge about one’s current location, destination, and the spatial relation between them: spatial knowledge. When people lack such knowledge, they become disoriented, if not totally lost. Disorientation can have serious consequences for people. It can lead to physical exhaustion, stress, anxiety and frustration, all of which threaten their sense of well being and limit their mobility (Bell et al., 1996; Carpman and Grant, 2002; Evans, 1980; Lynch, 1960). It may lead people to avoid or leave a place. In contrast, easy wayfinding may evoke positive feelings and a desire to visit. Arthur and Passini (1992) suggested the design process should include wayfinding requirements as an integral part. Planners and designers should produce “wayfinding plans” like

“HVAC plans.”

In the past few years, organizational

administrators have been considering wayfinding as a management issue and consulting professionals to improve wayfinding in their facilities (Carpman and Grant, 2002). Reduction in wayfinding problems can improve the image of a well-

1

maintained facility. It can also translate into dollars – benefits of increased productivity and reduced mobility costs. Easing wayfinding is particularly important for office buildings, airports, colleges, hospitals, libraries, museums, shopping malls, transit stations, entertainment parks, and zoos. Empirical studies can provide information that designers and planners can use to improve wayfinding. Toward this end, many studies attempted to understand the relation between several personal and physical environmental characteristics and spatial knowledge or wayfinding behavior (see reviews of Evans, 1980; Moore, 1979). Personal characteristics were studied extensively, but physical characteristics were rarely studied. Yet for planners and designers understanding the influence of physical environmental characteristics can help minimize the wayfinding difficulties. Moreover, many studies focus on one variable, but wayfinding involves many variables working simultaneously. My study explores the simultaneous effect of various personal and physical characteristics on wayfinding performance, and focuses more on the physical characteristics. Here I summarize the hypotheses, later I give a detail explanation (see chapter 2.3). For the personal characteristics, research showed age effect with better wayfinding performance for older children than younger children (Fenner et al., 2000; Heth et al., 1997) and for younger adults than older adults (Burns, 1998; Weber, 1978); gender effect with males having better wayfinding performance (Devlin and Bernstein, 1995, 1997; Lawton 1996); and familiarity effect with higher

2

familiarity producing better wayfinding performance (Ruddle et al., 199; Stanton et al., 1996, 1998). I hypothesized that Males and people with More Experience with the setting would show better wayfinding performance than Females and people with Less Experience with the setting. I expected that the performance would not be affected by age because my study targeted a narrow age group, young adults (18-55). I studied this age group as they represent the majority of the population. For the physical characteristics, research showed people tend to perceive wayfinding as difficult in complex layouts (Abu-Obeid, 1998; O’Neill 1991a; Weisman, 1981) and in environments with high degree of uniformity (lack of differentiation) (AbuGhazzeh, 1996; Passini et al. 2000). I hypothesized that environments with a Simple layout, higher Physical Differentiation, Vertical and Horizontal Differentiation should produce better wayfinding performance than environments with Complex layout, lower Physical Differentiation and with No Vertical or Horizontal Differentiation. In general, the findings supported the hypotheses. They suggest that to improve wayfinding, planners and designers should use simple layouts with some differentiation. Other researchers also suggested similar guidelines for planners and designers, however they rarely discussed how to manipulate plan layout complexity and physical differentiation and empirically tested the effect of physical factors on wayfinding behavior with considering other personal factors. This study showed ways to objectively measure and manipulate plan layout complexity and physical differentiation (see Chapter 3.2) for wayfinding before or after construction.

3

As a demonstration of a methodology to study wayfinding behavior, it also provides a model for future studies on spatial behavior. This study used virtual environments (VEs), computer generated three dimensional environments to simulate physical settings. VEs have been used in many other applications (described later) but have rarely been used in research exploring spatial behavior in different settings. This study showed that VEs could be extended to such applications. VEs allow control of physical characteristics and allow users to navigate as if they are in real environments. Chapter 2 reviews the research on wayfinding. It discusses the concepts and terms related to wayfinding. It reviews the personal and physical environmental characteristics related to wayfinding behavior, and it reviews methodological issues, the ways to simulate a physical setting and measure wayfinding behavior. Chapter 3 discusses the methods adopted in the present study. The study used virtual environments to simulate a residential neighborhood. To measure wayfinding behavior, it used multiple measures including pointing an invisible destination, finding the shortest route from one location to another and sketch mapping the experienced environment. Chapter 5 presents the results. Chapter 6 discusses the strengths and limitations of the present study and outlines areas for further research and for design and planning application.

4

CHAPTER 2 LITERATURE REVIEW

“Wayfinding” is a research area in “perception and cognition”, which is a subfield of “environmental psychology” (Figure 2.1) (Bell et al., 1996; Ittelson et al., 1974; Stokols and Altman, 1987). A broad field of enquiry, wayfinding encompasses a range of disciplines, such as psychology, geography, and planning (Foreman and Gillett, 1997).

Psychology

Sociology

Anthropology

Geography

Architecture



Planning

Environmental Psychology Environmental Stress

Hazard Perception

Weather, air pollution, noise



Perception and Cognition Wayfinding

Figure 2.1: The hierarchical structure of the study area, wayfinding.

5

Perhaps as a consequence of this diversity of disciplines, approaches to understand wayfinding vary. To clarify this study’s approach, this chapter first describes the concepts related to wayfinding, such as environmental perception and cognition, spatial knowledge, and cognitive map. Then it discusses the significance of wayfinding studies. Next, it explores the physical environmental and personal characteristics that affect wayfinding behavior. Then, it reviews the tools that have been used to simulate physical environment. Finally, it discusses the ways to measure wayfinding performance.

2.1 The Concepts Related to Wayfinding Perception refers to the experience of world, which happens in a moment of time and requires little or no information processing, while cognition refers to the comprehension of the environment that involves more information processing, and requires some mental activity (Bell et al., 1996; Heft, 1996; Nasar, 1998). Perception and cognition depend on one another (Moore and Golledge, 1976). Perception leads to a reconstruction of cognitive structures and is influenced by such structures. With perception and cognition, people develop spatial knowledge about the physical environment to maintain orientation and find their way from one location to another. Gardner (1983, 1993) argued that humans have multiple types of intelligence. Researchers studying wayfinding classified the spatial knowledge into

6

three types: (1) landmark knowledge, (2) route or procedural knowledge, and (3) survey or configurational knowledge (Belingard and Peruch, 2000; Pick and Lockman, 1981; Siegel and White, 1975; Throndyke and Hayes-Roth, 1982). Landmark knowledge refers to the concrete knowledge about places; route knowledge (or procedural knowledge) refers to the knowledge of routes that connect places; and survey knowledge (or configurational knowledge) refers to an integrated understanding of the layout of the space with the interrelationships of the elements contained therein. A number of researchers merged landmark and route knowledge into one category (route knowledge) yielding two types of spatial knowledge, route and survey knowledge (Abu-Obeid, 1998; Aginsky et al., 1997; Lawton, 1996; Rossano et al., 1999; Sholl et al., 2000; Witmer et al., 1996). Research differs on the sequence of the development of spatial knowledge. Some studies found that route knowledge precedes survey knowledge (Abu-Obeid, 1998; Belingard and Peruch, 2000; Hart and Moore, 1973; Lawton, 1996; Shemyakin, 1962; Siegel and White, 1975), while others found that survey knowledge developed first, (Hirtle and Hudson, 1991; Stevens and Coupe, 1978; Wilton, 1979), and still others found that people developed and access both types of knowledge simultaneously (Cole and Reid, 1998; Foley and Cohen, 1984; Lindberg and Garling, 1982; Taylor and Tversky, 1996). The conflicting results may be an artifact of the test situation. People may learn an environment, from a map or navigation. When learning it from a map, people tend to develop survey knowledge. However, when learning it from navigation, people tend to develop route knowledge (Rossano et al., 7

1999; Taylor and Tversky, 1996; Thorndyke and Hayes-Roth, 1982). The degree to which the setting has distinguishable landmarks or paths may also affect the type of spatial knowledge developed (Evans, 1980). The study of cognitive mapping (information about physical environment) is related to spatial knowledge, because it tells how people mentally represent the physical environment in their minds (Garling et al., 1984; Garling and Evans, 1991; Garling and Golledge, 1993; Moore and Golledge, 1976). Information about how people imagine the physical environment can be used to design, plan and manage environments that facilitate easier use and more satisfaction during navigation (Lynch, 1976; Moore and Golledge, 1976). Cognitive maps not only contain information about places and their spatial relationships, but also contain attributive values and meaning (Kitchin, 1994; Garling et al., 1984). Comparison of cognitive maps with cartographic maps found that similar to cartographic maps, cognitive maps are based on Euclidean geometry. People represent elevation differences in cognitive maps as such difference appear in cartographic maps. (Garling and Golledge, 1989). However, people’s cognitive representations are not perfect replicates of cartographic maps (Moore and Golledge, 1976). They are inaccurate and distorted (Downs and Stea, 1973; Kitchin, 1994; Moore and Golledge, 1976). The accuracy depends in part on the familiarity and experience with the environment (Garling et al., 1984; Golledge et al., 1985), and the

8

process through which a person acquires environmental knowledge, through map or navigation (Moore, 1979). Cognitive maps are important aids to wayfinding (Garling et al., 1984; Garling and Golledge, 1989; Passini, 1984a, 1984b). Wayfinding is defined as a behavior (Carpman and Grant, 2002) to reach a spatial destination or to navigate and orient in spatial environments (Devlin and Bernstein, 1995, 1997; Passini, 1984a, 1984b; Prestopnik and Roskos-Ewoldsen, 2000).

The wayfinding process is a kind of

problem solving. It involves 1) decision making, 2) decision execution, and 3) information processing (Passini, 1984a, 1984b; Passini et al., 2000).

2.2 Significance of the Wayfinding Research Wayfinding is prerequisite of satisfaction of other higher level goals (Weisman, 1981). It may not represent the primary performance goal but it certainly is necessary to perform tasks within an environment (Colle and Reid, 1998). Disorientation produces frustration, irritation, anxiety, and stress (Carpman and Grant, 2002; Evans, 1980; Lang 1987; Lawton, 1994). It can threaten our sense of well-being (Lynch, 1960), and limit personal mobility (Burns, 1998). Disorientation also has costs in terms of time and fuel, which contributes to congestion (Burns, 1998; Passini, 1980). In confusing places, such as hospitals and campuses, staff may waste time directing and leading people to locations (Hecht, 2000; Peponis et al.,

9

1990). Wayfinding difficulties may lead people to avoid places such as shopping malls, museums, and convention centers (Carpman and Grant, 2002). It can make people late for important occurrences such as business meetings or planes, which may cause loss of opportunity and money (Carpman and Grant, 2002). More serious consequences can result when ambulance drivers, firefighters have difficulty finding their way around (DeParle, 1989 as cited in Carpman and Grant, 2002). Although wayfinding has been a topic of interest for many disciplines including environmental psychology and geography, it is still not completely understood (Carpman and Grant, 2002). Organizational administrators, interior designers, architects, and planners can improve wayfinding when they understand how the physical environment affect wayfinding performance. The present study attempts to develop such an understanding. The next section discusses the physical environmental and personal characteristics that may affect wayfinding behavior.

2.3 The Physical Environmental and Personal Characteristics Affecting Wayfinding Behavior What physical environmental and personal characteristics may influence wayfinding behavior? First consider the physical environmental characteristics. My study considers four kinds of physical environmental features: (1) layout of the physical environment, (2) level of physical differentiation, (3) vertical differentiation

10

(presence of landmarks), and (4) horizontal differentiation (presence of road hierarchy).

2.3.1 The Plan Layout Researchers have suggested that the legibility or complexity of plan layout may effect wayfinding performance and cognitive mapping (Abu-Obeid, 1998; Appleyard, 1970; Garling et al., 1981; Garling and Golledge, 1989; Lynch, 1960; O’Neill, 1991a, 1991b; Passini, 1980; Weisman, 1981). Findings agree that people easily comprehend the physical environments if the plan layout is legible and simple. Weisman (1981) compared people’s self reports of wayfinding performance in a number of settings that vary in plan layout. People tended to perceive wayfinding as more difficult in settings that were more complex and less legible. O’Neill (1991a) found that people drew more accurate sketches and found their way to a specific destination more accurately in simple layouts. Researchers assessed the legibility or complexity of a plan layout through both subjective and objective measures. Weisman (1981) asked people to rate (subjectively) a sample of corridor diagrams according to degree of complexity. AbuObeid (1998) subjectively assigned three campus layouts to simple grid layout, semigrid layout, and compositive network. O’Neill (1991a) developed an objective measure of floor plan complexity which was based on the average number of paths at each decision point (see chapter 3 for a more detailed description). It gives designers

11

and planners specific information about what to manipulate to produce the desired effect. Thus, I used O’Neill’s objective measure. I expected the simple environments to produce better wayfinding performance than complex ones.

2.3.2 The Level of Physical Differentiation Higher physical differentiation may effect wayfinding behavior because it facilitates extracting and understanding of physical information (Abu-Ghazzeh, 1996; Abu-Obeid, 1998; Appleyard, 1969; Evans et al., 1982; Garling et al., 1986; Garling and Golledge 1989; Passini et al., 2000; Weisman, 1984). Passini et al. (2000) interviewed Alzheimer patients and nursing home staff to identify the architectural interior design features that caused wayfinding difficulties for patients. He found that monotony of architectural composition increased wayfinding difficulties. Abu-Obeid (1998) interviewed students from three universities with different campus designs. Students at campus with a repetitive design (i.e. lacking differentiation) showed poorer performance in sorting pictures to represent a route in the campus. AbuGhazzeh (1996) interviewed students to rank the physical setting variables that caused spatial orientation and wayfinding problems at campus. The results showed that high degree of uniformity (lack of differentiation) was the major factor in feeling lost or disoriented. No study tested the effect of the level of physical differentiation on wayfinding behavior in a controlled environment. The previous studies measured the

12

level of physical differentiation subjectively. For example Abu-Obeid (1998) assigned three different campus designs to low, moderate and high level of differentiation. However, each campus differs from one another in several other factors, such as the plan layout, size, or topography. Weisman (1981) suggested that physical differentiation refers to the extent to which one location looks different from others. Evans et al. (1982) suggested that physical differentiation can be achieved through changes. The present study achieved differentiation in two ways, vertical and horizontal changes. LOW differentiation environments had no vertical or horizontal changes, MODERATE differentiation environments had either vertical or horizontal changes, and HIGH differentiation environments had both vertical and horizontal changes. I expected that wayfinding performance would improve as physical differentiation increased from LOW to MODERATE to HIGH. Now consider vertical and horizontal differentiation.

2.3.3 The Vertical Differentiation Lynch and Rivkin (1976) and Wagner et al. (1981) demonstrated that when walking around people look at vertical elements such as buildings, and window displays. The permanent and distinctive vertical elements are remembered more (Appleyard, 1969; Evans et al., 1982; Lynch, 1960) and are important in navigation and orientation (Cornell et al., 1989; Evans, 1980; Evans, 1984a, 1984b; Heth et al., 1997; Passini, 1980; Ruddle et al., 1997; Tlauka and Wilson, 1994). Lynch (1960)

13

referred to such distinctive vertical elements as landmarks. Studies agree on the positive effects of landmarks on wayfinding. Cornell et al. (1989) found that landmarks played an important role in wayfinding success of 6 and 12 year old children. Heth et al. (1997) found that compared to 8-year-old children, 12-year-old children used landmarks more when recognizing the places while reversing a route. Doherty et al. (1989) found that children and adults showed better performance in scene recognition task when there were landmarks. Ruddle et al. (1997) found that people navigate more accurately in simulated environments that had landmarks than those without landmarks. Tlauka and Wilson (1994) argued that landmarks are helpful but not sufficient to successfully navigate from one location to another. My review of literature highlighted three important issues to consider in relation to landmarks: the type, the attributes and the location of landmarks. For the type, researchers refer to two types of landmarks, global (Darken and Sibert, 1993) and local (Heth et al., 1997). Local landmarks, such as a flower pot or a lamp, are visible within a restricted locality (Ruddle et al., 1998). Global landmarks, such as a mountain, are visible from far away and from many places (Ruddle et al., 1998). For the attributes, the most important attributes of buildings for landmark qualities include form, visibility (Appleyard, 1969) and uniqueness (Abu-Ghazzeh, 1996; Evans et al., 1982). For the location, researchers argued that landmarks are more effective when placed at transition points (Allen, 1982; Heth et al., 1997). My dissertation used local landmarks, because planners and designers can rarely manipulate global landmarks. I gave landmarks unique forms and I located them at 14

intersections. I expected wayfinding performance to improve with the presence of this vertical differentiation.

2.3.4 The Horizontal Differentiation Lynch and Rivkin (1976) and Wagner et al. (1981) demonstrated that when walking around, people look at the ground as well. People notice the differentiation on the pavement. However, no one has empirically tested if this horizontal differentiation enhances people’s wayfinding performance, as does the vertical differentiation. Road hierarchy is an important factor in determining the legibility of a physical environment (Lynch, 1960). It also produces horizontal differentiation and may enhance wayfinding performance. My dissertation used variation of road pavement and road width to produce road hierarchy and horizontal differentiation. I expected horizontal differentiation to produce better wayfinding performance. Now consider personal characteristics that may produce wayfinding performance differences. My dissertation considered three kinds of personal differences: (1) age, (2) gender, and (3) familiarity (experience). 2.3.5 Age Studies compared wayfinding abilities within children (Acredolo, 1977; Acredolo et al., 1975; Fenner et al., 2000; Heth et al., 1997; Piaget and Inhelder, 1967; Siegel et al., 1978), between children and adults (Bell, 2002, Cornell et al., 1989, 1992) and within adults (Burns, 1998; Evans et al., 1984a; Ohta and Kirasic

15

1983; Passini et al., 1990; Weber et al., 1978). Most studies found better wayfinding performance for older children than younger children (Fenner et al., 2000; Heth et al., 1997) and for younger adults than older adults (Burns, 1998; Weber et al., 1978). Fewer studies found no difference between younger and older children (Bell, 2002; Lehnung et al., 2001), or between younger and older adults (Brown et al, 1998). Heth et al. (1997) examined children’s place recognition. The results showed that older children (12 year-old) used more reliable, stationary landmarks, were more attentive to spatial relations between designated landmarks, and showed better performance in recognizing being on or off the route, than younger children (8 yearold). Fenner et al. (2000) examined children’s wayfinding behavior. Younger (5 and 6 years-old) children produced greater wayfinding errors, when replicating a route in forward and reverse directions, than older children (9 and 10 years-old). Lehnung et al. (2001) compared children’s understanding of spatial environment. 11-years-old children needed fewer trials to learn the sequence of landmarks than 5 and 7-yearsold ones, but there was no difference between ages 5 and 7 –years-old children. Older children also showed better performance in pointing landmarks than younger ones. Bell (2002) compared children’s and adults’ sketching ability. 7 and 9 year old children performed similarly in locating object, but adults performed better than the children. Weber et al. (1978) asked young adults to explore a nursing home for five minutes then compared their recognition of places with older residents. Younger adults recognized the places more accurately than older residents. Burns (1998) distributed a postal questionnaire survey, asking about wayfinding problems, to a 16

large sample of UK motorist drivers. Self-reports showed that older drivers tended to perceive wayfinding as being more difficult than non-elderly. Brown et al. (1998) asked people aged 20 to 78 to give directions to a stranger while looking at a map. When giving directions with access to a map, young and old adults used equal number of landmarks, road names relational turns and cardinal directions which were equally accurate. One explanation of contradicting findings on age relates to the test situation. Tests may differ in the load placed on memory. Older adults and younger children may show poorer performance on tasks that require memory. However, when memory demands are minimal or absent, the age effect on wayfinding behavior may disappear. For example, providing a map reduces the load on memory by eliminating the need for cognitive maps. Brown et al. (1998) found that when provided a map older and younger adults did not differ. Pointing with a compass puts a higher demand on memory than pointing with finger. Lehnung et al. (2001) found that 5year old group may show similar accuracy with 9 and 11-years-old children when they point with a finger but poorer performance when they point with a compass. Another explanation of contradicting findings may relate to variation in age breakpoints. Each study compared different age groups. This dissertation targeted only one age group, young adults. As a result of the narrow range of ages, I expected no effect of age on wayfinding performance.

17

2.3.6 Gender The effect of gender on wayfinding behavior is unclear in the literature. Some found gender differences in wayfinding tasks (Appleyard, 1976; Brown et al., 1998; Burns, 1998; Devlin and Bernstein, 1995, 1997; Evans, 1980; Galea and Kimura, 1993; Holding, 1989; Holding, 1992; McGuiness and Sparks, 1983, Miller and Santoni, 1986; Prestopnik and Roskos-Ewoldsen, 2000; Schmitz, 1997; Ward et al., 1986) and others found no such difference (Carr and Schissler, 1969; Cousins et al., 1983; Kirasic et al., 1984; Montello and Pick, 1993; Prestopnik and RoskosEwoldsen, 2000; Sadalla and Montello, 1989; Schmitz, 1997; Taylor and Tversky, 1992). Several studies find males performing better than females. Devlin and Bernstein (1995) had participants learn a campus with different cue sources (a series of photographs, a series of textual direction information, a map and combination of these cues). The participants were then shown another series of photographs and were asked to pick some to represent a route connecting A to B. Results showed that females had more errors than males. In a similar study, Devlin and Bernstein (1997) had participants study a neighborhood map. The participants were then asked to locate a number of landmarks on the map and indicate which path to follow to get from A to B. The results showed that males were faster than females in showing which path to follow. Lawton (1996) found gender difference in pointing task. He

18

gave participants a “landmark learning” tour in a building. After the tour, participants pointed to the landmarks at a location, which was not on the route they learnt. Results showed lower error scores for males. While these studies showed males performing better than females, many studies found no differences related to gender. Looking at a pointing task, Sadalla and Montello (1989) found no gender difference. In their study, participants walked a number of paths containing one turn. At the final point, participants pointed to the start point, the original direction of travel and also estimated the angle at the turn. Males and females performed similar in all three pointing tasks. Schmitz (1997) found some similarities and differences between females and males. He asked participants to explore a maze in forward and reverse directions. Then participants represented the maze in drawing or writing. Results showed that females were slower in exploring the maze than males. Females used more landmarks and fewer directions in their representations but overall females and males used similar number of elements to represent the maze. Prestopnik and Roskos-Ewoldsen (2000) employed a pointing task and a self-reporting test, related to sense of direction and use of wayfinding strategies. Self-reporting test showed no gender difference. Pointing task showed no gender difference in response time but male superiority in error scores. If there is a gender difference it may relate to biological factors, such as differences in brain organization (Kimura, 1992 as cited in Lawton 1996). However gender differences may also relate to other factors, which lead to contradictory

19

findings. First, researchers suggested that females depend more on route knowledge, while males depend more on survey knowledge (Lawton, 1994 and 1996). Using different types of wayfinding strategies (Galea and Kimura, 1993; Schmitz, 1997; Sholl et al., 2000; Ward, 1986) may cause differences in some tasks but not in others. If we accept the argument that route knowledge precedes survey knowledge (AbuObeid, 1998; Belingard and Peruch, 2000; Hart and Moore, 1973; Lawton, 1996; Shemyakin, 1962; Siegel and White, 1975) then we may expect no gender difference on tasks that require route knowledge but male superiority on tasks that require survey knowledge. If we accept the argument that route and survey knowledge develop simultaneously (Cole and Reid, 1998; Foley and Cohen, 1984; Lindberg and Garling, 1982; Taylor and Tversky, 1996;), we may expect female superiority on tasks that require route knowledge and male superiority on tasks that require survey knowledge. Second, studies showed that females tend to feel higher anxiety and fear (Devlin and Bernstein, 1997; Prestopnik and Roskos-Ewoldsen, 2000; Schmitz 1997), and less confidence (Lawton 1996) while completing wayfinding tasks, and they perceive wayfinding as more difficult (Burns, 1998). Such feelings may lead to poorer wayfinding tasks performance (Evans et al., 1984b; Garling and Golledge, 1989). Third, in some cultures males have greater opportunity to travel and thus develop better directional skills (Evans, 1980).

In such cultures, the greater

experience for males may give them better wayfinding scores than females.

20

Despite the contradicting findings on gender, in this study I expected male superiority in spatial knowledge because I used tests of survey knowledge to measure it (see section 2.5).

2.3.7 Familiarity (Experience) In general, studies found positive effects of experience on spatial knowledge or wayfinding performance (Ruddle et al, 1998; Stanton et al., 1996, 1998). Ruddle et al. (1998) did a controlled study in a simulated environment. They had participants explore a simulated environment repeatedly. The number of times a participant navigated in the environment determined the level of familiarity. The spatial knowledge tasks included finding a route, estimating direction and distance. Participants developed more accurate spatial knowledge with increased familiarity. Stanton et al. (1996, 1998) had children explore simulated environments at fortnightly intervals. Children’s performance on spatial tasks, such as estimating direction and sketching (drawing an outline of the environment, placing the objects on a prepared outline plan) improved with repeated exploration, especially for large scale environments. In this dissertation, I expected that wayfinding performance would improve with increases in physical environment experience. 2.3.8 Summary of Factors Effecting Wayfinding Performance I have shown that wayfinding may relate to personal characteristics of the wayfinder and the physical characteristics of the environment. Personal

21

characteristics include age, gender, familiarity (experience). Physical characteristics include plan layout, physical differentiation and its components vertical and horizontal differentiation. Most studies looked at the effect of each factor alone. To better understand wayfinding, we need to consider personal and environmental factors simultaneously. In two early review papers, Moore (1979) and Evans (1980) concluded that although personal factors were widely explored, physical environmental factors were understudied. My review of the literature shows a continuing insufficient attention to the influence of physical environmental characteristics on wayfinding behavior. Hence, this study focused on the physical characteristics, but considered them and personal characteristics simultaneously. It did the tests in controlled conditions. To study wayfinding in a controlled physical setting, researchers must decide on at least two kinds of factors, 1) the ways to simulate the environment, 2) the ways to measure wayfinding responses. I used Virtual Environments to simulate the environment and multiple measures of survey knowledge to test wayfinding performance. The following sections discuss my choices and the other possible options to simulate a physical setting and measure wayfinding performance.

2.4 Tools to Simulate Environment Researchers have observed and tested people’s wayfinding behavior in real or simulated settings. Real environments have variety that is hard to control. Manipulating the structure of real environment is difficult or even impossible

22

(Belingard and Peruch, 2000). However, to test the effect of specific physical environmental factors, one needs to control the variety in the physical setting. Controlling the physical structure is possible in simulated settings. Simulation, in this study, refers to visually representing something real. Visual simulations include: (1) photographs, (2) small-scale three dimensional models, (3) full-scale three dimensional models, (4) three dimensional computer models. Each tool varies in level of detail, color, quality of reproduction, and conditions in which the simulation is depicted, and in which it is seen (Sanoff, 1991).

2.4.1 Use of Photographs and Simulation Booth Photographs may simulate on-site experience and have been widely used in environmental psychology, in studies of environmental preference (Nasar, 1988), perception (Appleyard et al., 1964; Bosselman 1998) and wayfinding (e.g. O’Neill, 1991a). A perceiver traveling along a route perceives sequence of transitions that connect successive vistas (Cullen, 1961; Heft, 1996). This sequence of transitional information is particularly important for wayfinding behavior. However, pictures do not show continuous movement of transitions. Heft and Nasar (2000) questioned the static nature of photographs because they lack the dynamic experience of moving through places. Their comparisons of people responses to dynamic displays (videotaped segments along a route), and static displays (freeze frames of each segment) found differences between two display modes.

23

Reseachers have presented a series of photographs to simulate movement in a drive (Appleyard et al., 1964) or walk (Bosselman, 1998; Cullen, 1961; O’Neill, 1991a). A series pictures would feel like watching a movie. However, the slides in such simulations restrict the observer’s ability to scan the environment extensively (Sanoff, 1991). Possible views that one can see in a real setting are not shown in these sequential photos (Bosselman, 1998). An alternative way to simulate movement through photographs is using a simulation booth. A simulation booth has screens encircling the viewer to give three dimensional view of an environment. It also provides opportunity for active exploration. Winkel and Sasanoff (1966) (as cited in Sanoff 1991) compared the movement pattern of people in a real environment and in a simulation booth. The results confirm the similarities between path selection in each condition, but interviews after the simulated trip revealed that people might not have acquired a comprehensive image of the space in the simulated environment.

24

2.4.2 Full Scale Models Full scale models have been used to assess a proposed designs (Marans, 1993; Sanoff, 1991), to understand people’s perception of an existing environment (Sadalla and Oxley, 1984) and to observe people’s wayfinding behavior (Passini et al., 1990; Schmitz, 1997, Waller et al. 1998). Marans (1993) built a model of a hospital room. Different user groups (nurses, doctors and visitors) explored this model and assessed the use quality prior to construction. Sadalla and Oxley (1984) built rooms of the same size but different shapes and asked people to estimate their size to see if shape effected perception of size. Passini et al. (1990) built a maze and asked people to do several wayfinding tasks. A full scale model provides opportunity to control factors that vary in the real environment while letting people move as if they are in a real setting. However full scale models may look like a maze, and it requires large spaces, time and money (Figure 2.2). With full scale models, one can simulate an indoor space but not a large outdoor space (Sanoff, 1991).

25

Figure 2.2: Full scale models (Sanoff 1991)

2.4.3 Small Scale Models Building and modifying small scale models are relatively easy and less costly than full-scale models. However, their static nature and inability to allow views to experience movement at the eye level represent major drawbacks that may differentiate them from experience in full scale places (Sanoff, 1991). To simulate movement researchers have put a small camera in a model. A camera moving around a small-scale model records the scenes on a videotape transmits the visual information to a television screen. Movement direction and speed of the camera can be controlled (Figure 2.3).

26

Figure 2.3: The left one is the simulator at Lund (Sanoff 1991; p 146), right one is the simulator at the Institute of Urban and Regional Development at the University of California at Berkley. (Altman and Wohlwill, 1977; p.81)

Such simulators have been used to understand and foresee the impact of plans (Appleyard and Craik - as cited in Bosselmann 1998; Janssens and Kuller, 1986 - as cited in Sanoff, 1991) and to predict the people’s behavior (Carpman et al., 1985). Appleyard and Craik had people evaluate a setting in three ways; an actual tour by driving through the real setting in van, a video recorded in the real setting and a video recorded using a model to simulate driving. The results demonstrated that people found the simulated environment realistic. Carpman et al. (1985) built a small scale model of a hospital setting with two different design, having a parking deck accessed directly from a drop-off entrance drive or having a parking deck accessible only from the main road. In both designs a sign is leading people to access the parking deck through the main road. They simulated a drive by recording a videotape in each design. Visitors were shown these simulations and were asked where they would turn to park if they were coming alone to visit a patient. Results showed that in the design with a direct access from the drop-off circle, people tended to ignore the sign and

27

chose to turn to the drop-off circle to enter the parking deck because they see the entrance located adjacent to the drop-off circle. With this simulation Carpman et al. (1985) could observe people’s behavior to two different designs before construction, and could see the potential problem of crowdedness at the drop-off circle if there is an entrance to the parking structure from it. With such simulations, respondents reported that they felt like exploring a real environment. However, such simulators were rarely used as a research tool, perhaps because of its high costs (Bosselman 1998; Carpman et al. 1985; Sanoff, 1991)

2.4.4 Computer Models (Virtual Environments) Computer generated three-dimensional environments are called virtual environments (VEs). In these simulations, the user can visualize and interact with the virtual three-dimensional spatial environment in real time. VEs are used in research related to physical environment to control the physical characteristics (Arthur et al., 1997; Rossano et al., 1999; Wilson et al., 1997a) or when it is hard to gather subjects in the real one (Ishikawa et al., 1998). Because this technology is relatively new1, there are some concerns about its limitations related to its cost, its ability to represent real world experience and its ability to transfer spatial information to real world experience.

1

The ideas behind this technology began to emerge about 20-30 years ago (see Kalawsky, 1993; Rehingold, 1991, for reviews, as cited in Wilson, 1997. 28

For cost, it varies because the hardware and software requirements vary. Until recently the necessary hardware and software was limited to high end technologies; however, recent improvements have lowered costs enough to make it available in many offices and have improved the image quality enough to make it an efficient tool for environmental research (Aginsky et al., 1997; Garling et al., 1997; Ishikawa et al., 1998; Peruch 1998; Rossano et al., 1999; Wilson et al., 1997a). For similarity between the real and VE experience, studies explored if people could develop spatial representations in VEs and if people’s navigation behavior in VE is similar to the one in the real environments. Studies consistently showed that people developed good spatial representation of VEs (Regian et al., 1992; Peruch et al., 1995). Although, one study showed that people developed route knowledge slower in a VE than in an equivalent real environment (Witmer et al., 1996), others showed that people who navigated in a virtual environments showed similar accuracy in spatial knowledge tests to that of people navigated an equivalent real-world building (Rassona et al. 1999; Ruddle et al, 1997, 1998; Throndyke and Hayes- Roth, 1982; Wilson et al., 1996). For transfer of spatial information from VEs to real environments, studies tested if people could perform some spatial tasks in the real environment by just exploring a virtual simulation of the real environment. Studies showed that people did transform the spatial information they gained while exploring a virtual replica of the real environment (Bliss et al., 1997; Cromby et al., 1996a, 1996b; Wilson et al. 1996).

29

For example, Wilson et al. (1996) found that students who explored a virtual building containing only the major features of real building, such as corridors and doors, could estimate directions to the objects in the real environment. Cromby et al. (1996b) found that children who explored a virtual supermarket were able to find the items in the real supermarket. Bliss et al. (1997) found that firefighters who learned a building from a simulation were able to find their direction in the real building.

2.4.5 Summary of Tools for Studying Wayfinding This study used simulated environments rather than real environments because it allowed control of environmental factors. From the possible simulation tools, this study used computer generated models for three reasons: (1) they are dynamic and active where people can navigate as if they are in real environment; (2) they are flexible and easy to manipulate or control the physical features of the environment; and (3) they are affordable.

30

2.5 Measures of Wayfinding Performance Researchers have measured wayfinding performance in five ways: 1) self report tests, 2) memory tests, 3) recognition tests, 4) spatial orientation tests, and 5) navigation tests.

2.5.1 Self Report Tests Self report tests include tasks such as reporting subjective judgments of wayfinding ability or navigational strategy. Weisman (1981) and Abu-Ghazzeh (1996) developed a questionnaire to measure respondent’s own judgments of wayfinding ability. It asked about wayfinding behavior (e.g. have you ever had trouble finding your way?), and perceived knowledge and understanding of the setting (e.g. how confident would you be of the directions you’d give to such a stranger?). Lawton (1994, 1996) developed a questionnaire to measure respondents’ navigational strategy while finding their way around an unfamiliar building. The questionnaire listed possible strategies, and the respondents rated the likelihood of using each strategy on a 5-point scale. Some strategies related to route knowledge, such as using room numbers and signs, others related to survey knowledge, such as using global reference points. Prestopnik and Roskos-Ewoldsen (2000) used Lawton’s (1994) questionnaire. Dogu and Erkip (2000) combined Weisman (1981) and Lawton (1996) questionnaire to examine people wayfinding behavior in a shopping mall.

31

Long questionnaires are not the only way to measure self-judgements of wayfinding behavior. Researchers also asked wayfinders to rate their sense of direction (Scholl 1988; Kozlowski

and Bryant, 1977; Prestopnik and Roskos-

Ewoldsen, 2000) and to tell about their wayfinding decisions (Murokashi and Kawai, 2000; Passini, 1984b). Respondents gave reasons for their choice of route whenever there were choices. Self-report tests do not require much effort, time and money. However they have been criticized for lack of consistency with the actual behavior (Ericson and Simon, 1984; Marlowe et al., 1965 as cited in Judd et al., 1991). People tend to exaggerate or give socially desirable answers (Lam and Cheng, 2002).

2.5.2 Memory Tests Memory tests include tasks such as describing places or routes after a trip. For example Appleyard (1969), Carr and Schissler (1969), and Lynch and Rivkin (1976) had people tell which points or places they remember best after a trip, not necessarily in an order. Such a test tells only about people’s knowledge about the presence of particular places. Wayfinders may need this information to verify being on the right trail, but they also need to know the connection between and sequence important places. As a result some researchers had people describe a route between any given two points or tell what they had seen in sequential order after the trip (Appleyard, 1969; Carr and Schissler, 1969).

32

Memory tests are inexpensive and easy to administer, but they may reflect the individual’s language ability rather than spatial knowledge. An individual may only report recalling places that are easy to describe in words. One study compared what people really looked at (items recorded with a head-mounted Polymetric eyemovement recorder) with what people remembered looking at (Carr and Schissler, 1969). It found agreement, but people tended to report items that were easy to name even though they had not look at them.

2.5.3 Recognition Tests Recognition tests include tasks such as naming the objects during a trip, recognizing a scene with pictures after the trip and sorting the pictures to show the route. For example, Brunswik (1944) and Wagner et al. (1981) followed people and stopped them at some intervals to ask what they were looking at. Magliano et al. (1995), Aginsky et al. (1997), Wilson (1999) and Murakoshi and Kawai (2000) showed participants pictures from the test environment and distracter pictures similar to the ones in the test environment but from different locations that the participant had never seen. They asked the participants to tell whether they had seen it in the environment or not. Heth et al. (1997) escorted children from an origin to a destination (original route) and then from the destination to the origin (return route), but the return route had loop branches attached to the original route. On the return route, they stopped children at some intervals and asked whether they were on or off

33

the original path. Such tests measure if people recognize being somewhere when they are actually there. People may recognize being there but not know what they will see next. To test such knowledge, Magliano et al. (1995) showed pairs of pictures from the route and asked participants to decide which of the two pictures came first along the route. Abu-Obeid (1998) asked students to arrange a series of pictures to show a route. Like memory tests, recognition tests are inexpensive and easy to administer. Recognition tests measured during a trip may not reflect the real life experience because the participant would attend the surrounding environment more in the test situation than in the real life (Lynch and Rivkin, 1976). Also, tests with pictures may be biased when the picture is taken from a different perspective than the observer’s perspective. Both memory and recognition tests measure wayfinding performance indirectly. They tell if an individual remembers a certain place and has information about a route between those places, but they do not reveal the individual’s knowledge about how those places are spatially related.

2.5.4 Spatial Orientation Tests Spatial orientation tests include three kinds of tasks: (1) drawing a sketch map, (2) estimating distances between visible or invisible locations, and (3) estimating directions of invisible locations. First consider sketch maps. Kitchin

34

(1997) discussed five variations of sketching: (1) the basic sketch map technique, where the researcher gives the respondent a blank sheet of paper to sketch a map. (2) the normal sketch mapping technique, where the researcher imposes constraints to obtain required data (3) the cued sketch mapping technique, where the researcher gives a portion of the map and asks the respondent to complete the specific features, (4) the longitudinal sketch mapping technique, where the researcher asks respondents to draw the map on layers of carbon tracing paper and turn the sheets over at some time intervals to study how sketch map evolves, and (5) the cloze sketch mapping technique, where the researcher covers a base map in a grid with some squares deleted and has respondents fill the information in the blank squares. Appleyard (1969) and Kitchin (1997) asked residents to draw a map of the city on a blank paper. Carr and Schissler (1969), Murakoshi and Kawai (2000), and Aginsky et al. (1997) showed participants a route and asked them to draw the route on a blank sheet with or without commenting on the location and characteristics of the various elements. Schmitz (1997) had participants explore a maze repeatedly and ask them to represent the maze graphically. O’Neill (1991a) showed a series of pictures depicting paths through a building, showing decision, starting and destination points, then asked people to draw a sketch map of the floor plan and mark the location of start and destination points. Wilson (1999) provided a plan view of the environment and asked participants to name locations and indicate the position on the plan. Rossano et al. (1999) laid out a number of white post-board cut-out shapes on a desk, and asked participants to pick a shape as the experimenter reads name of a building 35

from the learned environment. Then the participants were asked to arrange the selected shapes on a blank sheet. Rossano and Reardon (1999) gave participants a sheet on which two buildings in the environment was represented with a rectangle and asked them to locate the third building. Kitchin (1997) gave a map which had blank boxes and asked residents to fill the blank boxes with the name of landmarks provided in a list. Sketch maps may reflect the individual’s drawing ability more than spatial knowledge. However, one study found a weak correlation between the accuracy of sketch map and artistic ability (Rothwell, 1976 as cited by Evans, 1980). Sketch maps may also have metric and topological distortions. People tend to draw turns as right angle (even if they are not), enlarge the areas that include lots of turns at the expense of straight roads, distort the relative length of road segments (Aginsky et al., 1997). Interpreting the accuracy of a sketch map is challenging. Passini (1984b) argued that the sketches should be assessed according to their utility value instead of their cartographic value. Researchers used different approaches to measure the accuracy of sketches. They rated the overall accuracy2 (Aginsky et al., 1997; Murakoshi and Kawai, 2000;) and complexity (Appleyard, 1970; Rovine and Weisman, 1989), judged the accuracy of placement of path segments, location of intersection choice (O’Neill, 1991a), counted the frequency of path, node, landmarks, and turns (Aginsky et al., 1997; Rovine and Schmitz, 1997; Weisman, 1989), calculated the difference 2

Two researchers rated the maps independently. Maps with nearly correct placements of the route were rated 4; maps that were toplogically correct with some distortions were rated 3; maps with almost complete elements but with toplogically incorrect layouts were rated 2; and fragmental maps or blank maps were rated 1. 36

between correct and incorrect turns (Aginsky et al., 1997), measured the relative distance between landmark pairs (Aginsky et al., 1997), and assessed the topological accuracy considering if the landmark was in the appropriate sequence with respect to tour and the path between landmarks showed the appropriate turns (Rovine and Weisman, 1989) or considering the number of breaks in the map (Aginsky et al., 1997). No one measure is better than the others. Thus a combination of such measures would give better measure. Using sketch maps, researchers also calculated distance estimates (Biel, 1982; Jansen-Osman and Berendt 2002; Rossano et al., 1999; Wilson, 1999) and direction estimates (Colle and Reid, 1998; Gillian, 1994; Rossano et al., 1999; Tlauka and Wilson 1996) between landmarks. Now consider distance estimations. Researchers used many tasks (in addition to sketching) including verbal estimation, drawing straight lines, reproducing a walk or comparing the route choices (see Montello, 1991 for a review). Belingard and Peruch (2000) asked participants to estimate the straight line distance between the current and target locations. Peruch et al. (1989) asked general public and taxi drivers to estimate straight line distance and travel distance verbally in either distance or time units. Biel (1982) asked children to decide which one of the two landmarks were closer to the current location. Jansen–Osman and Berendt (2002) had people explore three routes. Given a straight line representing the length of one of the three routes, the participants were asked to mark the length of the other two routes. Sadalla and Magel (1980) and Sherman et al. (1979) had participants re-walk a straight line distance at the same length as the test distance. Nasar (1983) examined the estimated 37

distance to places by asking which of the two parking garages people chose to park to minimize the distance to their office. Nasar (1985) also had people estimate walking and crow-flies distance to hidden and visible destinations. Distance estimate is easy to measure. The performance in distance estimates were usualy calculated as the difference between the true distance and the estimated distance. Finally consider direction estimates. Researchers used two kinds of pointing task (in addition to sketching): (1) simulated pointing tests and (2) actual pointing test (Throndyke and Hayes-Roth, 1982; Rossano et al., 1999). In simulated test, participants were asked to imagine their current location and position, and then point to a target location. In actual test, people were asked to point to a target location when they were in the real environment. Again, direction estimate is easy to measure. The performance was measured as time to respond, with quick response indicating better performance. The accuracy was calculated as the difference between the true direction and the estimated direction. People may point to a target by changing their looking direction, drawing on a paper, using a keypad or a handheld pointer. Belingard and Peruch (2000) requested participants to rotate their view to face an invisible target. Murakoshi and Kawai (2000) told people to show direction of a target with an arrow. Rossano et al. (1999) asked participants to mark along the perimeter of the circle to show the direction to the target after explaining that the center of the circle represents the current location and an arrow extending up from the circle represents the looking direction.

38

Prestopnik and Roskos-Ewoldsen (2000) explained participants that numbers on the keypad represent the 450 angles surrounding them. Participants were then asked to pick a number to show if the target is in front, back, left, right, front left, front right, back left, or back right of them. Rossano et al. (1999) Sholl et al. (2000), and Lawton (1996) used a hand held pointer. The pointer is mounted within a circular dial. Participants rotate the pointer arm to show the direction of the target. Spatial orientation tests measures how people represent spatial environment but not how they navigate. Anooshian (1996) argued that ‘people often report knowing the identities and locations of landmarks while having little or no knowledge of how to navigate from one landmark to another.

2.5.5 Navigation Tests Navigation test include tasks such as finding places, replicating a route, reversing a route, finding the shortest path between two places and describing a route to a stranger. Abu-Ghazzeh (1996) asked people to find some destinations in three different buildings that they had never visited. Peponis et al. (1990) had participants freely explore the setting and they had them search for specific locations. Rossano et al. (1999) asked people to lead the way from one building to another in a campus after studying a map of the setting or exploring the computer simulation of the setting. O’Neill (1991a) asked participants to replicate the route they learned with a series of pictures. Fenner et al. (2000) guided children along a route in a campus then

39

asked them to walk along the route without assistance in forward and reverse directions. Schmitz (1997) had people find their way in a maze first from origin to destination then from destination to origin. Murakoshi and Kawai (2000) guided people to a destination with some detours then asked them to go back to origin by using the shortest possible route. Rovine and Weisman (1989) took people on a tour and pointed out some important target locations throughout the tour, then asked participants to select the most direct route from a target to another one. Ward et al. (1986) and Brown et al. (1998) asked people to give directions to a hypothetical stranger while looking at the map or after memorizing the map. Passini et al. (1990) did a series of navigation tests. First, they guided people in a maze, after which they asked them to retrace the path on their own. Second, they showed people two routes which were intersecting, and asked them to retrace a combined route (the first half of the first trip and second half of the second route). Third, they had them learn a route from a small scale-model, and then had them execute a route in the maze. Finally, they showed a route and asked participants to find a shortcut back to the departure. Navigation tests show people’s wayfinding behavior more accurately than self reporting tests, but they might be time consuming. Accuracy in navigation test can be measured in many ways. Researchers recorded the success in taking the most direct route (Rovine and Weisman, 1989), amount of time spent to complete the task (O’Neill, 1991a; Murakoshi and Kawai, 2000), the route and the distance (AbuGhazzeh, 1996; Murakoshi and Kawai, 2000; Rovine and Weisman, 1989), the speed (Schmitz, 1997) number of backtrackings (Abu-Ghazzeh, 1996; O’Neill, 1991a), 40

number of turns (Rovine and Weisman, 1989), number of wrong turns (Murakoshi and Kawai, 2000; O’Neill, 1991a; Rossano et al. 1999), and number of times people asked for direction or used maps (Abu-Ghazzeh, 1996; Rossano et al. 1999).

2.5.6 Summary of Wayfinding Measures There are numerous tests to measure an individual’s wayfinding performance. Each test has a different requirement which causes variation in the performance. For example, people may find their way accurately, but they may not be able to draw comparable sketch maps3 (Clayton and Woodyard, 1981; Downs and Siegel, 1981; Hart 1981; Passini 1984b). Passini (1984b) explained this contradiction with a different cognitive memory requirement. The cognitive memory requirement is higher for sketching than route finding, because route finding requires recognition while sketching requires recall. Individuals recognize (notice) places they encounter. However, they recall (remember) places by retrieving the environmental information stored previously. The contradiction occurs not only between the tests but also between the tasks (within each test). For each test, different tasks may measure different spatial knowledge. For example, for spatial orientation tests, estimating travel distances or directions between successive locations measures route knowledge, while estimating crowflies distances and directions measures survey knowledge (Goldin and

3

Note, this differentiation may apply to recognition and memory tests. 41

Throndyke, 1982; Golledge, 1987). For navigation tests, reproducing a route represents route knowledge, while taking a shortcut represents survey knowledge (Loomis et al., 1993; Presson and Montello, 1994). Any single measure may have a bias. By using distinct and different measures one can reduce the effects of each bias (Judd et al., 1991; Kitchin, 1996). In this study, I used multiple measures of survey knowledge to understand wayfinding performance. I asked people to report the strategy they used (self report), point to an unseen place (spatial orientation test- estimating direction task), pick the correct map among possible ones and mark some locations on that map (spatial orientation testsketching task) and find the shortest route between two locations (navigation testfinding the shortest route task).

42

CHAPTER 3 METHODOLOGY 3.1 General Procedures and Equipment Interviews took place in five university dormitories at The Ohio State University; Morill Tower, Steeb Hall, Park Hall, Stradley Hall and Jones Tower, all but Jones Tower house undergraduate students. The experimental session lasted between five to ten minutes per participant. The tests took place on weekdays and weekends, from 9:30 am to 7:30pm, in January 2003. The completion of 166 tests took 70 hours.

3.1.1 Introductory Procedures The study met the university’s human subjects requirements. Posters informed passers-by that by taking a five minute computer-based survey they would help a Ph.D. student and receive a cookie and a soft drink. The participants received a brief written description about the study (Appendix A). The description indicated that the participant could withdraw at any time without penalty. 43

3.1.2 Equipment and Setting A graphic PC-based desktop workstation (Pentium II, 32MB graphics card, resolution 640X480X256, 17 inch monitor) was set up on a desk in a quiet location, close to the entrance in each building. I faced the computer screen to a wall so participants could not see the simulated environment before the test. Participants were seated facing 50 cm from the center of the screen, to achieve a visual field of approximately 45 degrees. A camcorder, placed on the same desk with the desktop workstation, videotaped the computer screen to track the navigation.

3.2 Virtual Environments

3.2.1 Software The simulated environments were built using a three-dimensional modeling program, GTK Radiant. A real-time three-dimensional environment generator game engine, QUAKE III ARENA software, produced perspective views to simulate ground level walk-paced movement through the simulated environment. The viewpoint was set at a height of 5.6 feet, average eye level. The game engine displayed the scenes in color at a rate of approximately 20 frames per second. Participants controlled their movement in the simulated environment via keyboard. It provided left right rotations (left right arrows) and forward backward translations (up-

44

down arrows), with movement restricted to the horizontal plane. Other researchers who used the keyboard for controlling motion demonstrated that users easily mastered this form of interaction (Tlauka and Wilson, 1994; 1996; Wilson et al. 1996, 1997; Jansen-Osman and Berendt, 2002; Belingard and Peruch, 2000). The direction of the observer’s gaze paralleled their direction of movement (i.e., no side view was available).

3.2.2 Physical Environmental Characteristics I created eighteen simulated environments. The environments contained houses, trees, sidewalks, open green spaces, parking and background skyline to represent relevant aspects of a residential area. I replicated a two-story house plan, surrounded with a tree, a street lamp and a parking space (Figure 3.1). The environments contained repeated units of this house plan. Repeating a design is a common planning approach for some real developments and apartment complexes. This study attempted to simulate such environments because residents and visitors of those uniform developments or apartment complexes often have difficulties to find their way. I derived the texture maps from digital photographs of real buildings and real objects and overlaid these textures onto the modeled objects to achieve detail and realism. Vivid colors, real-world textures, and visual elements gave a strong impression of a real residential setting.

45

House

Tree Lamp Sidewalk Road

Figure 3.1: The same house plan was repeated in all environments. From top left moving clockwise images show plan view (not seen by subjects), front view, right view and left view.

A collisions-detection algorithm was used to prevent walking through walls and to constrain walking to the roads. At each intersection choice point a message reminded the participants that they need to choose the next direction. Arrows placed at the intersections indicated the possible directions the participant could choose (Figure 3.2).

46

Figure 3.2: The arrows at intersections showed possible directions one can take and a message reminds users that they can change direction.

Each model differed on three physical factors: the plan layout (simple or complex), vertical differentiation (without landmark, with landmark type A-object or landmark type B-building), and horizontal differentiation (without road hierarchy, with road hierarchy type A-road width variation or road hierarchy type B-road pavement variation). I classified the eighteen environments into three groups according to the level of physical differentiation, the extent to which parts of environment looks different from others (Table 3.1). The “low differentiation” group included the environments without any landmark or road hierarchy. The “moderate differentiation” group included the environments with either a landmark (one of two kinds) or road hierarchy (one of two kinds). The “high differentiation” group included the environments with both a landmark (one of two kinds) and road hierarchy (one of two kinds).

47

Table 3.1: Level of differentiation was determined by the presence of vertical and horizontal differentiation

Level of Differentiation * Low Differentiation Moderate Differentiation High Differentiation

With Vertical Differentiation (Landmark) No Yes (one of two kinds) No Yes (one of two kinds)

With Horizontal Differentiation (Road Hierarchy) No No Yes (one of two kinds) Yes (one of two kinds)

* Same set of conditions for simple and complex environments

For plan layout, half of the environments had a complex layout and half had a simple layout. I used O’Neill’s (1991a) “Inter Connection Density” (ICD) measure to determine simple and complex plan layout. The measure is based on the density of interconnections at choice points. Figure 3.3 shows an example of the calculation of ICD from a plan. The number of connections at each intersection, or choice point, is listed to the right of the plan (ie. at intersection A, one has two choices). ICD is calculated as the mean number of connections. Hence the plan has an ICD of 2.33.

A

B

C

D

A=2 B=2 C=3 D=3 E=2 F=2 +______ 14

F

E

14/6 = 2.33 = mean ICD

Figure 3.3: An example for calculating Interconnection density (ICD) value

48

Using this measure, O’Neill (1991a) identified three settings inside the State University of New York at Buffalo Lockwood Library, with different ICD measures to examine the relationship between plan layout complexity and wayfinding performance. Figure 3.4 shows the plan layouts and the schematic drawings of the simplest and the most complex settings used in that study. In the schematic drawings, you can see that the simple setting had ten intersections, and the complex one had eleven intersections (the small and large circles). Both settings had five choice points (the large circles), one START point, and one DESTINATION point. The simple setting had an ICD of 2.4 and the complex setting had an ICD of 2.54.

Complex

Schematic Drawing

Plan Layout

Simple

Destination

LEGEND

Destination

Choice Points Intersections Start

START and DESTINATION

Start

Figure 3.4: The plan layout and schematic drawings of the simple and complex settings in O’Neill’s study.

49

My study replicated O’Neill’s (1991a) simplest and most complex plan layouts for outdoor virtual environments (Figure 3.5). My environments comprised a rectangular area of 312 x 273 meters (1,024 x 896 feet). The complex plan layout contained 27,017 meters (88,640 feet) of road, and the simple plan layout contained 24,189 meters (79,360 feet) of road. As a collusion detection algorithm prevented navigation through some roads, the complex plan had 8,993 meters (29,504 feet) of walkable road, and the simple one had 6,574 meters (21,568 feet). Thus, simple plan had 73% of walkable road of the complex plan.

Complex

Plan Layout

Simple

Schematic

LEGEND Roads

Market

Market

Walkable Roads START and MARKET

Start

Start

Choice Points

Figure 3.5: The plan layout and schematic drawings of the simple and complex settings in this study.

50

To represent the start and destination points from O’Neill’s (1991a) study, the virtual environments had two signs, START and MARKET (Figure 3.6).

Figure 3. 6: The START and MARKET signs in all environments.

For vertical differentiation, the environments differed according to the presence of landmarks. One third of the environments did not have landmarks while two thirds did. To give the results greater generalizability there were two types of landmarks. Half of the environments “with landmark” had object landmarks (type A), such as a lamp, a flag, and a flowerpot (Figure 3.7). The other half had building landmarks (type B) differing in color and shape from one another and from the other buildings (Figure 3.8). Each type (A and B) had four landmarks.

51

Figure 3.7: Environments with vertical differentiation had two types of landmarks. TYPE A had four object landmarks shown from top left moving clockwise (one kind of lamp, another kind of lamp, a flower pot and a flag) at choice points. (All environments were in full color).

52

Figure 3.8: Environments with vertical differentiation had two types of landmarks. TYPE B had four building landmarks that differ from one another and the surrounding buildings shown from top left moving clockwise (a gray brick building, an orange brick building, a white building and a yellow building) at choice points. (All environments were in full color).

Because landmarks are more effective when they are located at changes in course or direction of traveling (Allen, 1982, Abu-Obeid, 1998), I located them at choice points (Figure 3.9).

53

LEGEND Roads Walkable Roads START and Market Signs Landmarks Figure 3.9: The location of landmarks in the Simple and Complex environments.

For horizontal differentiation, the environments differed according to the presence of road hierarchy. One third of the environments did not have road hierarchy while two thirds did. To give the results greater generalizability, I used two types of road hierarchy. In half of the environments “with road hierarchy,” I varied road width (Type A, narrow or wide) to create road hierarchy. In the other half, I varied the pavement (Type B, asphalt or cobblestone) to create road hierarchy (Figure 3.11). In environments with no road hierarchy, all roads had the same width (wide) and pavement (asphalt). In environments with road hierarchy, the most efficient route between START and MARKET signs was wide or had asphalt pavement and all other roads were narrow or had cobblestone pavement. Figure 3.10 shows the hiearchy of roads in the Simple and Complex plans.

54

LEGEND Roads (narrow or cobbelstone) Roads (wide or asphalt) Walkable Roads START and Market Figure 3.10: In environments with road hierarchy the most efficient route between START and MARKET signs were wide or had asphalt pavement and all other roads were narrow or had cobblestone pavement.

55

Type A (Width differentiation)

Type B (Pavement differentiation)

Perspective view at intersection

Perspective view at intersection

Perspective view at wide road

Perspective view at asphalt road

Perspective view at narrow road

Perspective view at cobblestone road

Figure 3.11: Environments with horizontal differentiation had two types of road hierarchy. For TYPE A (left column) road width varied and for TYPE B (right column) road pavement varied.

56

3.2.3 Realism of Virtual Environments Judgment Participants rated the simulated environments as realistic (M = 4.87, SD = 1.43, where 1=not realistic at all, 7=very realistic, n=160). Most respondents rated the realism above average (63%). Some rated it as average (20%) and fewer rated it as below average (17%). Participants who gave low scores often said that the simulation was realistic but what was simulated was not realistic. Although some found the computer graphics quality very high, they gave low realism scores because of the difference between their own neighborhood and the simulated environment. They said their neighborhood was more spacious and had more variation in landscape and housing types compared to the simulated environment. I examined if judged realism differs across different personal and physical environmental groups. Table 3.2 shows the means for the judged realism across different personal groups. The judged realism did not differ at a significant level across gender (F (1,158) = 0.027) or computer game playing frequency (F (6,153) = 0.690).

57

Table 3.2: The means of judged realism across personal characteristics

Judged Realism Mean (SD)

Personal Characteristics Gender Female (n=65) Male (n=95) Computer game playing frequency Never 1 (n=23) 2 (n=33) 3 (n=27) 4 (n=15) 5 (n=31) 6 (n=14) All the time 7 (n=17)

4.85 (1.28) 4.88 (1.54) 5.13 (1.40) 4.55 (1.25) 4.93 (1.03) 5.27 (1.53) 4.84 (1.51) 5.00 (1.66) 4.65 (1.94)

Table 3.3 shows the means for the judged realism across different physical environmental groups. The judged realism did not differ at a significant level across plan layout (F (1,158) = 0.027), level of physical differentiation (F (2,157) = 1.638), vertical (F (1,158) = 3.367) and horizontal hierarchy (F (1,158) = 0.513), the type of landmark (F (2,157) = 2.127) or road hierarchy (F (2,157) = 0.693).

58

Table 3.3: The means of judged realism across physical environmental characteristics

Judged Realism Mean (SD)

Physical Characteristics Level of Physical Differentiation Low Differentiation (n=40) Moderate Differentiation (n=80) High Differentiation (n=40) Plan Layout Complex (n=80) Simple (n=80) Vertical Differentiation Without Landmark (n=80) With Landmark (n=80) Landmark Type A (n=40) Landmark Type B (n=40) Horizontal Differentiation Without Hierarchy (n=80) With Hierarchy (n=80) Road Hierarchy Type A (n=40) Road Hierarchy Type B (n=40)

4.60 (1.48) 4.85 (1.48) 5.18 (1.25) 4.89 (1.39) 4.85 (1.48) 4.66 (1.42) 5.08 (1.42) 4.93 (1.42) 5.23 (1.42) 4.79 (1.53) 4.95 (1.33) 4.80 (1.39) 5.10 (1.28)

I also examined the first order correlation between judged realism and computer game frequency. They did not have a statistically significant Pearson correlation (r=-.016, p>.05)

59

3.3 Participants and Group Demographics 166 volunteers (98 male, 68 female) participated in the study. Most participants (about 85%) were students in a range of programs at the Ohio State University and a small portion (about 15%) were staff. All had normal or corrected to normal vision. I dropped 6 participants from the sample because they did not complete the whole survey. For the remaining 160 volunteers (95 male, 65 female) the ages ranged from 18 to 48, but most participants (83%, 133 people) were under 25. Only 9 participant (6%) were older than 30. When asked how often they played computer games the participants rated themselves on averages as playing between rarely and sometimes (M=3.67, SD=1.92 where 1=never, 7=all the time). Studies found that males are more likely to play computer games (Greenfield, Brannon, and Lohr, 1994; Greenfiled et al., 1994; Phillips, Rolls, Rouse, and Griffiths, 1995; Subrahmanyam and Greenfiled, 1994; Colwell, Grady, and Rhaiti, 1995). My study paralleled those results. A statistically significant difference for gender in reported computer game playing frequency (F(1,158) = 49, p