Identifying and Measuring Urban Design Qualities Related to Walkability

Final Report prepared for the Active Living Research Program of the Robert Wood Johnson Foundation

July 2005 Measuring Urban Design Qualities Related to Walkability

Final Report prepared for the Active Living Research Program of the Robert Wood Johnson Foundation

Reid Ewing and Otto Clemente National Center for Smart Growth Research and Education University of Maryland

Susan Handy Department of Environmental Science and Policy University of California, Davis

Ross C. Brownson Prevention Research Center Saint Louis University School of Public Health

Emily Winston Institute of Transportation Studies University of California, Davis

June 2005

ii

Table of Contents 1. Introduction ...... 1 Conceptual Framework ...... 1 Work Plan ...... 2 2. Recruit Expert Panel...... 3 Panel Selection...... 3 Panel Members...... 3 3. Identify Key Urban Design Qualities...... 4 Initial Screening of Qualities...... 4 Selected Qualities ...... 6 Challenge ...... 7 4. Acquire Videotaping Equipment ...... 7 5. Develop Filming Protocol...... 7 6. Create a Library of Video Clips and Select a Sample ...... 8 Fractional Factorial Design ...... 8 7. Administer Visual Assessment Survey to Expert Panel ...... 11 Survey Protocol ...... 11 8. Establish Criteria for Selecting Urban Design Qualities to Operationalize...... 13 9. Analyze Relationships from the Visual Assessment Surveys ...... 13 Correlates of Overall Walkability ...... 13 Inter-rater Reliability of Scene Ratings ...... 14 10. Analyze Content of Sampled Scenes ...... 15 Inter-rater Reliability of Content Analysis...... 15 11. Relate Urban Design Ratings to Measurable Physical Features...... 20 Cross-Classified Random Effects Models...... 20 Results of Statistical Analysis ...... 22 Discussion ...... 22 12. Select Which Qualities to Define Operationally...... 28 13. Develop a Draft Instrument for Measuring Urban Design Qualities ...... 29 14. Test and Refine the Draft Instrument...... 30 15. Train Lay Observers in the Classroom ...... 33 16. Field Test the Survey Instrument with Lay Observers ...... 34 17. Finalize the Instrument ...... 36 Appendix 1: Biosketches of Expert Panel Members ...... 33 Appendix 2: Qualitative Definitions of Urban Design Qualities ...... 40 Imageability...... 40 Legibility...... 40 Enclosure...... 42 Human Scale...... 43 Transparency ...... 44 Linkage ...... 44 Complexity ...... 45 Coherence ...... 47 Appendix 3: Operational Definitions of Physical Features ...... 49 Appendix 4: Relationships Between Urban Design Qualities and Physical Features ...... 68 Appendix 5: References...... 73

iii Identifying and mediate) between the physical This Round I Active Living Research features of the environment and Measuring Urban Project has sought to develop walking behavior. Physical features of Design Qualities operational definitions and the built environment influence the measurement protocols for nine quality of the walking environment Related to Walkability intangible qualities of the urban both directly and indirectly through 1. Introduction environment. It has been largely the perceptions and sensitivities of successful and forms the basis for individuals. Our study focuses on A growing body of research provides further work. Operational definitions urban design qualities, qualities of evidence of a link between the built and measurement protocols have the environment that depend on environment and active living. been developed for six of nine physical features but reflect the However, to date, the measures used perceptual qualities, specifically: general way in which people perceive to characterize the built environment imageability, visual enclosure, human and interact with the environment. have been mostly gross qualities such scale, transparency, complexity, and as neighborhood density and street tidiness. These perceptual qualities are connectivity (see reviews by Ewing different from qualities such as sense and Cervero 2001; Handy 2004; and To aid in the dissemination of the of comfort, sense of safety, level of Ewing 2005). measures, we have developed a field interest that reflect how an individual survey instrument. This instrument reacts to a place–how they assess the The urban design literature points to will be tested and refined, and lay conditions there, given their own subtler qualities that may influence observers will be trained in its use. preferences and perspectives. choices about active travel and active With valid and reliable measures Perceptions are just that, leisure time. These qualities will be available, and relatively easy to use, perceptions. They may produce referred to as perceptual qualities of travel behavior researchers and different reactions in different people. the urban environment or, physical activity researchers will be They may be assessed with a degree alternately, just as urban design better able to study relationships of objectivity by outside observers. qualities. The urban design literature between the built environment and presumes that these qualities are walking behavior. All of these factors—physical important for walkability, without features, perceptual qualities, and Conceptual Framework much empirical evidence. Until urban individual reactions—influence the design qualities can be measured, The conceptual model underlying this way that an individual feels about the this presumption will remain study considers the role of environment as a place to walk. By untested. perceptions as they intervene (or measuring these intervening

1 variables, we can better understand the relationship between physical features of the built environment and walking behavior. Work Plan Our work plan consisted of 16 tasks:

• Recruit expert panel. • Identify key urban design qualities. • Acquire videotaping equipment. • Develop filming protocol. • Create library of video clips and select a sample. • Administer visual assessment survey to expert panel. • Establish criteria for selecting which urban design qualities to operationalize. • Analyze relationships from the visual assessment survey. • Analyze content of sampled scenes. Figure 1: Conceptual Framework • Relate urban design quality ratings to measured physical features. • Select urban design qualities to define operationally. • Develop a draft instrument for operationalizing urban design qualities. • Test and refine the draft instrument.

2 • Train and certify lay observers principal investigators (PIs) called paradigm) and those with a more in classroom. colleagues and asked for conventional orientation toward • Provide field training in use of nominations. Experts identified in urban design. the instrument. this manner were asked for additional Panel Members • Finalize the instrument. nominations, and so the process proceeded until all 10 openings were The 10 expert panel members Tasks are described in subsequent filled. selected through this process were: sections of this report. The criteria for panel membership Victor Dover - urban designer – were as follows. First and foremost, Principal-in-Charge, Dover, Kohl & 2. Recruit Expert Panel panel members had to have Partners Town Planning, South Miami A critical part of our workplan was to knowledge of urban design concepts FL assemble a panel of urban design and and expertise in urban design or planning experts. The panel related fields that would support their Geoffrey Ferrell - urban members helped define perceptual subjective judgments. designer/code expert - Geoffrey qualities of urban scenes, rated Second, so their opinions would carry Ferrell Associates, Washington, D.C. different scenes with respect to these weight, the panel members had to qualities in a visual assessment-style acknowledged leaders in their Mark Francis - landscape architect – survey, submitted to interviews as respective fields. This is an Professor, Department of Landscape they assigned their ratings to provide exceptionally creative and prominent Architecture (check on this), the research team with qualitative group. University of California, Davis, CA insights into their rating criteria, met to discuss ways of measuring Finally, the panel members had to Michael Kwartler - perceptual qualities, and reviewed represent a variety of different architect/simulations expert – and commented on the draft field perspectives. Having both planning Director, Environmental Simulation observation manual which presented and public health represented among Center, New York, NY the measurement instrument in all its the PIs ensured that different detail. Their views on the character professional networks would be Rob Lane - urban designer - Director and importance of different urban tapped. An effort was made to of the Design Program, Regional Plan design qualities became the gold achieve a balance between urban Association, New York, NY standard for this study. designers and other planning professionals, and between those Anne Vernez Moudon - urban Panel Selection with a “new urbanist” bent (referring designer/planner - Professor of The panel was selected through a to a movement in design and Architecture, Landscape Architecture, networking process in which the planning that has its own design and Urban Design & Planning,

3 University of Washington, Seattle, WA

Tony Nelessen - urban designer – President, A. Nelessen Associates, Inc., Princeton, NJ and Associate Professor of Urban Planning, Rutgers University, New Brunswick, NJ

John Peponis - architect/space syntax expert - Associate Professor of Architecture, College of Architecture, Georgia Institute of Technology, Atlanta, GA

Michael Southworth - urban designer - Professor, Department of City and Regional Planning and Department of Landscape Architecture, University of California, Berkeley, CA

Dan Stokols - social ecologist – Professor, School of Social Ecology, University of California, Irvine, CA Figure 2: Visual Assessment Survey During Expert Panel Meeting Biosketches of expert panel members are contained in Appendix 1. activities takes place, and where one street more walkable than 3. Identify Key Urban walking for exercise mostly occurs. another. Parks, plazas, trails, and other public Design Qualities Initial Screening of Qualities places have a role in physical activity, In terms of the public realm, no but due to budget and time Key perceptual qualities of the urban element of the urban environment is limitations, this study deals only with environment were identified based on more important than streets. This is the perceptual qualities that make a review of the urban design where active travel to work, shop, eat literature. At different points in the out, and engage in other daily

4 careers, two of the PIs have had environments and better understand occasion to review the urban design Lynch, Kevin. 1960. The Image of the what individuals value in their literature (see Ewing, 1996 and City. Joint Center for Urban Studies, environments. Partial listing of this Handy, 1992). Perceptual qualities Cambridge, MA. voluminous empirical literature is figure prominently in such classics provided in Ewing (2000) and as: Rapoport, Amos. 1990. History and updated in Ewing et al. (in press). Precedent in Environmental Design. These literature reviews go beyond Alexander, C., S. Ishikawa, and M. Kluwer Academic Publishers, Plenum the boundaries of urban design to the Silverstein. 1977. A Pattern Language Press, New York. fields of architecture, landscape - Towns Buildings Construction. architecture, park planning, Oxford University Press, New York. Sitte, Camillo. 1889. City Planning environmental psychology, etc., as According to Artistic Principles. Verlag perceptual qualities of the Arnold, Henry. 1993. Trees in Urban von Carl Graeser, Vienna. (Complete environment figure prominently in Design, Van Nostrand Reinhold, New translation in G.R. Collins, C.C. these fields as well. York. Collins, Camillo Sitte.1986. The Birth of Modern City Planning, Rizzoli The long list of perceptual qualities Cullen, Gordon. 1961. The Concise International Publications, New York). described in literature includes: Townscape. Reed Educational and adaptability, ambiguity, centrality, Professional Publishing, London. Trancik, Roger. 1986. Finding Lost clarity, compatibility, comfort, Space—Theories of Urban Design. complementarity, continuity, Gehl, Jan. 1987. Life Between Van Nostrand Reinhold, New York. contrast, deflection, depth, Buildings - Using Public Space. Van distinctiveness, diversity, dominance, Nostand Reinhold, New York. Unwin, Raymond.1909. Town expectancy, focality, formality, Planning in Practice. T. Fisher Unwin, identifiability, intelligibility, interest, Hedman, Richard. 1984. Ltd., London. (Reissued by Princeton intimacy, intricacy, meaning, Fundamentals of Urban Design. Architectural Press, New York, 1994). mystery, naturalness , novelty, American Planning Association, openness, ornateness, prospect, Chicago, IL. Whyte, William H. 1988. City— refuge, regularity, rhythm, richness, Rediscovering the Center, Doubleday, sensuousness, singularity, Jacobs, Allan. 1993. Great Streets. New York. spaciousness, territoriality, texture, MIT Press, Cambridge, MA. unity, upkeep, variety, visibility, and The research team has also reviewed vividness. Jacobs, Jane. 1961. The Death and the visual assessment literature, Life of Great American Cities. which attempts to measure how We focused on eight urban design Random House, New York. individuals perceive their qualities that appeared distinct and

5 important from both the qualitative Legibility Transparency literature and quantitative attempts Legibility refers to the ease with Transparency refers to the degree to to measure what is valued by users which the spatial structure of a place which people can see or perceive of urban space: imageability, can be understood and navigated as what lies beyond the edge of a street legibility, visual enclosure, human a whole. The legibility of a place is or other public space and, more scale, transparency, linkage, improved by a street or pedestrian specifically, the degree to which complexity, coherence. The eight are network that provides travelers with people can see or perceive human defined in the next section. A ninth a sense of orientation and relative activity beyond the edge of a street quality, tidiness, was added when a location and by physical elements or other public space. Physical review of video clips indicated that that serve as reference points. elements that influence transparency one potentially important dimension include walls, windows, doors, of scenes simply was not captured by Enclosure fences, landscaping, and openings the eight qualities selected initially. Enclosure refers to the degree to into midblock spaces. Selected Qualities which streets and other public spaces Linkage are visually defined by buildings, A definitional piece was written on walls, trees, and other elements. Linkage refers to physical and visual perceptual qualities of the urban Spaces where the height of vertical connections from building to street, environment. Definitions were elements is proportionally related to building to building, space to space, refined, and references were added, the width of the space between them or one side of the street to the other with help from the expert panel. have a room-like quality. which tend to unify disparate Short definitions are supplied below. elements. Tree lines, building Longer discussions are provided in Human Scale projections, marked crossings all Appendix 2. Human scale refers to a size, texture, create linkage. Linkage can occur

and articulation of physical elements longitudinally along a street or Imageability that match the size and proportions laterally across a street. of humans and, equally important, Imageability is the quality of a place Complexity correspond to the speed at which that makes it distinct, recognizable, humans walk. Building details, Complexity refers to the visual and memorable. A place has high pavement texture, street trees, and richness of a place. The complexity imageability when specific physical street furniture are all physical of a place depends on the variety of elements and their arrangement elements contributing to human the physical environment, specifically capture attention, evoke feelings, and scale. the numbers and kinds of buildings, create a lasting impression. architectural diversity and ornamentation, landscape elements,

6 street furniture, signage, and human image stabilizer; and extended second clip – about 40 seconds – activity. battery life. slow 180 degree pan and back again

Coherence Three identical video camcorders third clip – about 20 seconds – slow Coherence refers to a sense of visual were acquired so the video clips could 360 degree pan order. The degree of coherence is be shot concurrently by the three co- influenced by consistency and PIs and research assistants in their fourth clip – about 55 seconds - slow complementarity in the scale, distant geographic settings. Detailed 360 degree pan – slow walk forward character, and arrangement of specifications for the chosen – slow 360 degree pan buildings, landscaping, street camcorder can be found at furniture, paving materials, and other http://www.canondv.com/zr70mc/ind fifth clip – about 55 seconds – slow physical elements. ex.html. 360 degree horizontal pan with slow vertical pans in each direction Challenge 5. Develop Filming Protocol Our challenge was to move from With five individuals shooting clips, a We quickly realized that much of the these highly subjective definitions of consistent filming protocol had to be environment relevant to a pedestrian urban design qualities to operational followed to ensure that reactions to was not captured by stationary clips. definitions which capture the essence street scenes were not biased by Slow forward motion became of each quality and can be measured different filming techniques. A great standard despite the wobble this with a degree of reliability across deal of experimentation and dialogue occasioned. The image stabilizer was raters, including those without among the PIs went into the of only assistance. Both continuous training in urban design. development of a protocol that would and intermittent panning horizontally were tested. The continuous pan 4. Acquire Videotaping mimic the experience of pedestrians. Pedestrians are usually in motion, covered more ground in a given time. Equipment sway a bit as they walk, have Vertical panning was added to After a survey of equipment options, peripheral vision, and tend to scan approximate peripheral vision and we settled on the digital video their environments. capture vertical elements that might camcorder because it would allow attract attention. extended video clips and easy For example, one set of early trials downloads to a computer. The Canon compared: Filming was extended for longer ZR70 MC Mini DV Camcorder was periods so as to take in more of the selected for its moderate price and first clip – about 60 seconds – fast street. The beginning of the block specific features: ability to shoot 180 degree pan and back again – was established as the consistent both video and stills; wide angle slow walk forward – fast 180 degree starting point. We considered the attachment; precision optical zoom; pan and back again possibility of continuing to film until

7 the end of the block became visible - looking straight ahead, pan down travels of PIs on other business but realized this would consume too 30 degrees, pan up 30 degrees, back during the course of the study. much time on longer blocks. At some to level point, the wide pan to the rear was While they succeed to varying dropped as disorienting. The full pan - pan 45 degrees right, pan up to the degrees, all streets included in the to the right was also dropped since it top of adjacent buildings or trees, sample attempt to accommodate brought buildings uncomfortably and back to level pedestrians. All could be close. characterized as urban. All have - pan 135 degrees left to the sidewalks. All offer pedestrian The final filming protocol became: opposite side of the street, pan up to amenities of some sort such as the top of opposite buildings or trees, landscaping, pedestrian lighting, Have video camcorder set on high and back to level street furniture, and trip destinations resolution. Using the "Canon Wide within view. Attachment WA-30.5" lens and - pan 90 degrees right, to straight Fractional Factorial Design zooming out as far as possible ahead manually, proceed as follows: Scenes were shot and ultimately - repeat the sequence for a total of selected for the visual assessment - start about 20 feet from the two complete pans survey using a fractional factorial beginning of the block on the outside design. We wanted to capture of the sidewalk 6. Create a Library of Video relevant combinations of the eight Clips and Select a Sample urban design qualities being - walk slowly forward in the direction Working off a shoot list, more than operationalized (tidiness was added of adjacent traffic at a speed of 200 clips were filmed in dozens of later). Without variation across the approximately 1 mph cities around the United States. The qualities, it would be impossible to shoot list was generated according to tease out the contributions of - as you move forward, pan slowly a fractional factorial design described physical features to the urban design and continuously following the same momentarily. Diversity of street quality ratings of our expert panel. sequence scenes was ensured by the different regional settings of the PIs, and the

8 A common experimental design is Table 1: 28-4 Fractional Factorial Design That Served as a Guide for Sampling Video Clips one in which all input factors are set at two levels each. In the literature, these levels are referred to as `high' Human and `low' or `+1' and `-1', Imageability Enclosure Scale Transparency Linkage Complexity Coherence Tidiness respectively. A design with all clip low low low low low low low low possible high/low combinations of all 1 the input factors is called a full clip high low low low low high high high factorial design in two levels. 2 If there are k factors, each at 2 clip low high low low high low high high 3 levels, a full factorial design requires 2k separate runs (in this case, the clip high high low low high high low low 4 “runs” are individual video clips). clip low low high low high high high low Even if the number of factors, k, in a 5 design is small, the 2k runs specified clip high low high low high low low high for a full factorial can quickly become 6 very large. With two levels and eight clip low high high low low high low high factors, a full factorial design requires 7 256 runs. clip high high high low low low high low 8 The solution to this problem is to use clip low low low high high high low high only a fraction of the runs specified 9 by the full factorial design. A clip high low low high high low high low fractional factorial design is 10 considered a better choice when clip low high low high low high high low 11 there are five or more factors, as there were in this study. A fractional clip high high low high low low low high 12 factorial design is one in which only a clip low low high high low low high high fraction of the treatment 13 combinations required for the full clip high low high high low high low low factorial experiment is selected. That 14 fraction may be ½, ¼, etc. of the clip low high high high high low low low runs called for by the full factorial. 15 Properly chosen fractional factorial clip high high high high high high high high designs for 2-level experiments have 16

9 perfectly. Others matched on only seven, six, or even five of the qualities, rather than all eight. Urban design qualities tend to co-vary (that is, appear in certain combinations of high and low values), making perfect matches unlikely starting with any practically sized set of clips. These carefully selected clips served as our sample in the subsequent visual assessment survey. The 28-3 sample allowed us to capture the main effects of each urban design quality on overall walkability, plus two-factor interaction effects.

To expand the sample size for analytical purposes, 16 additional clips were later selected. For this sample, we sought to match the 28-4 fractional factorial design (as shown in Table 1).

To illustrate, we found a clip with high values of all eight urban design qualities that perfectly matched the Figure 3: Best Match for a Run that Required High Values of All Eight Qualities corresponding 28-4 run (clip 16 in Table 1—shown in Figure 3). But we had to settle for a clip that matched the desirable properties of being both qualities (tidiness not included). values of only seven qualities for the balanced and orthogonal. From the larger set, 32 clips were 28-4 run that required high values of selected that best matched the imageability, human scale, linkage To choose our samples, one PI and combinations of high/low values in a 8-3 and tidiness and low values of other his RA rated clips as “high” or “low” 2 fractional factorial design. Some qualities (clip 6—shown in Figure 4). with respect to the eight perceptual of clips matched high/low patterns We, again, also settled for a clip that

10 matched only seven qualities for the run that is basically the opposite to clip 6 with high values for enclosure, transparency, complexity and coherence (clip 11—shown in Figure 5). Although we weren’t able to exactly match the fractional factorial design in all cases, following the design as close as possible resulted in the selection of clips that are distinctly different as the following figures illustrate. Where ratings for Figure 4: Best Match for a Run that Figure 5: Best Match for a Run that two or more clips matched factorial Required High Values of Imageability, Required High Values of Enclosure, design equally well, clips were Linkage, and Tidiness Transparency, and Complexity selected to maximize geographic diversity. scores, and the team member members who could not attend the 7. Administer Visual recorded scores and taped meeting were sent DVDs and comments. Thus, there was a subsequently surveyed by phone. Assessment Survey to quantitative and qualitative element Survey Protocol Expert Panel to the survey. The qualitative element would assist the research The three PIs and their RAs divided The first wave of the visual team in identifying physical features up responsibility for conducting assessment survey (32 clips) was of scenes worth measuring in the phone surveys with the expert panel conducted electronically. The sample subsequent content analysis, and members. Therefore, as with the of video clips was recorded in random would provide a fallback for shooting of video clips, it was order onto DVDs, and the DVDs were operationalizing urban design necessary to establish a standard distributed to expert panel members. qualities if the quantitative analysis protocol for the interviews so as to A telephone survey was then failed. ensure consistency and avoid conducted in which the panel possible bias in responses. Some member and a research team To expand the sample, a second experimentation was involved with member viewed each clip visual assessment survey (16 this protocol as well. concurrently, the panel member additional clips) was conducted face- assigned scores to each clip and to-face at a meeting of the expert The final protocol was as follows: commented on the specific features panel. For this survey, clips were of scenes that produced high or low also in random order. Panel

11 Before the interviews, contact recorder produces adequate sound allow them to rate with one decimal panelists and ask them to review the quality. Monitor the tape recorder place precision on overall walkability. work plan for the project. Also ask throughout the interview to make Make sure they provide reasons, them to have printed copies of the sure it does not run out of tape. It articulate criteria, and/or define survey form and perceptual qualities did so twice in our pilot session. relevant physical features for each definitional piece available during the rating. One key reason per rating will survey. These should be on their To show panelists the range of values be sufficient. desks during the survey, the former represented by the sample (so they to be completed and the latter as a leave room for ouliers at the top and Concurrently, you and the panelists reference. Panelists will be using bottom of the Likert scale), view clips will play clips in the order they their personal computers to view the 3, 7, 15, 16 before starting the rating appear on the DVD, which was DVD, and so will need hard copies of process. Mention that a consistent randomized. Move through the clips the survey form on which to record filming protocol was used throughout at panelists' desired pace. Since their ratings. which, we hope, gives them a their comments are being tape complete picture of the streetscape. recorded, it will not be necessary to Begin interviews by asking if Ask them if they have any questions take notes on qualitative reactions to panelists have any questions or or feedback on the clips. clips. But ask panelists to give you feedback on the purpose of the their quantitative ratings on the project, purpose of the visual Make sure panelists have the sound second pass through each clip. You assessment survey, the survey on their computers turned on for the will be recording their ratings on your instrument itself, or the urban design interviews. Ratings can and should survey form, as they concurrently qualities definitional piece. We have be affected by sound as well as sight record ratings on their survey forms. added still photos of scenes to the (so in this sense, this isn't a pure rating form. With the photos as visual assessment survey). When all 32 clips have been rated, memory jogs, panelists can refer Explain to the panelists that each clip ask them to go back on their own to back to earlier clips for benchmarks will be played two or more times. finalize their ratings in light of the as they proceed through the survey. Have them use the first pass to entire sample. Ask them to email the We have also added a final column to familiarize themselves with the scene final rating form to you in any event, the form, in which panelists will rate and comment on the streetscape in to check that you have correctly the overall quality of the walking open-ended fashion. On subsequent recorded ratings given over the environment for each clip. passes, have them assign phone. If every panelist grades quantitative ratings to all perceptual harder on later clips, we can control Conduct interviews on the speaker qualities on a 1-5 Likert scale. Ask for "order of viewing effects" phone, recording the entirety for later them to rate in whole numbers on the statistically. If not, we will have to reference. An inexpensive tape individual perceptual qualities, but rely on their own adjustments upon

12 re-viewing to achieve intra-rater (2) The urban design quality was Correlates of Overall Walkability reliability. rated by the expert panel with at Using mean values for the 48 video least a moderate degree of inter-rater clips, we found that overall These are exceptionally reliability (ICC > 0.4). walkability is directly and significantly knowledgeable people who will related to each urban design quality recognize many of the streets. You (3) The total variance in ratings of individually. The analysis is can confirm but don't identify places the urban design quality was complicated, however, by the fact so as not to bias ratings with positive explained to at least a moderate that eight of the nine qualities (the or negative associations. degree in terms of measurable exception being tidiness) are physical features of scenes (explained collinear. Tolerance values were 8. Establish Criteria for portion > 0.3). unacceptably low when all variables Selecting Urban Design were included in a regression at once. (4) The portion of total variance Qualities to Operationalize attributable to scenes was explained We realized early on that not all Linkage and legibility appeared to be to a substantial degree in terms of largely functions of the other urban urban design qualities could be physical features of scenes (explained defined operationally for design qualities, so they were portion > 0.6). dropped from further consideration. streetscapes. Appendix 2 includes a discussion of measurement Of the remaining variables, human (5) All physical features related to possibilities for the first eight scale had the strongest relationship ratings of a particular urban design to overall walkability almost qualities. Some are clearly more quality were measured with at least a amenable to measurement than are regardless of what combination of moderate degree of inter-rater variables was tested. Tidiness, and others. reliability (ICC > 0.4), excluding to a lesser extent, transparency, those for which ICC values could not enclosure, and imageability, were To decide which urban design be computed. qualities were ultimately defined somewhat independent of human operationally, five criteria were 9. Analyze Relationships scale, proved significant at the 0.10 established: from the Visual Assessment level in most model runs, and improved the explanatory power of Surveys (1) The urban design quality was the model (the adjusted R-squared). judged by the expert panel to have a Results of the visual assessment Coherence was ultimately dropped statistically significant relationship to surveys were pooled, and analyses because it proved insignificant and overall walkability ratings (p < 0.05). were performed with expert panel reduced the explanatory power of the ratings in connection with criteria (1) model. Complexity was ultimately and (2) of the preceding section. dropped even though significant in some model runs, because it altered

13 relationships between other variables Table 2: Regression Model for Overall Walkability and overall walkability, and because it had a low tolerance value itself. Standardized Variable Coefficient Coefficient t-statistic p-value The best-fit equation is presented in constant -0.226 -1.503 0.140 Table 2. Perceptual qualities explain human scale 0.411 0.420 5.814 0.000 more than 95 percent of the variation transparency 0.137 0.149 2.366 0.023 in mean overall walkability, according tidiness 0.070 0.059 1.598 0.117 to our expert panel. All qualities are enclosure 0.140 0.157 2.504 0.016 directly related to overall walkability, imageability 0.307 0.310 5.153 0.000 and all are significant at conventional N 48 levels except tidiness, which falls just R-square .959 Adjusted R- below the 0.10 level. Based on their .954 t-statistics, human scale ranks first in square significance as a determinant of overall walkability, imageability Table 3: Inter-rater Reliability for Ratings of Perceptual Qualities second, enclosure third, transparency fourth, and tidiness a distant fifth. Intra-class 95% Inter-rater Reliability of Scene Correlation Confidence Cronbach’s Ratings Coefficient Interval of ICC Alpha imageability .494 .385-.618 .930 Various statistical techniques may be legibility .380 .276-.509 .895 used to assess inter-rater reliability in enclosure .584 .478-.697 .945 studies like this, where multiple individuals independently rate the human scale .508 .399-.630 .928 same set of cases. For assessing transparency .499 .390-.622 .926 inter-rater agreement, intra-class linkage .344 .169-.621 .896 correlation coefficients (ICCs) are complexity .508 .398-.632 .926 more appropriate than simple coherence .374 .271-.504 .880 correlation coefficients. Simple tidiness .421 .314-.550 .915 correlation coefficients are sensitive N 48 only to random error (chance factors), while ICCs are sensitive to if two experts rate a group of scenes the other (systematic error), a both random error and systematic and one of them always assigns simple correlation coefficient would error (statistical bias). For example, scores that are x points higher than indicate complete agreement

14 between them. By contrast, the ICC physical features of scenes. The The gold standard for this activity would accurately portray the extent procedural alternative to this was provided by one of the PIs and of disagreement between them. The approach, giving users either his RA, who developed detailed ICC is the preferred measure of inter- qualitative criteria or pictorial operational rules for measuring each rater reliability when cases are rated examples upon which to based urban physical feature. The features and in terms of some interval variable or design ratings, seemed fraught with operational definitions are listed in interval-like variable, such as the subjectivity and imprecision. Appendix 3. The process might best Likert scales used in this study. be described as one of forced Toward this end, all 48 video clips consensus. The PI and RA In this study, the 10 expert panelists were analyzed for content. Physical independently measured each independently rated each of 48 clips features of each scene were feature, discussed differences, and with respect to the nine urban design “eyeballed” and then quantified with finally reached agreement on a single qualities, and values were compared as much care and precision as the value for each physical feature of for inter-rater reliability (see Table medium allowed. All told, more than each video clip. 3).From their ICC values, most urban 100 features were measured in this design qualities demonstrate manner for each scene. The process Inter-rater Reliability of Content moderate inter-rater reliability among typically required more than an hour Analysis panelists (0.6 > ICCs > 0.4); the for each video clip, and much more exceptions—linkage, coherence, and for the more complex scenes. Just how reliably physical features legibility—show fair reliability (0.4 > can be measured is a criterion in the ICCs > 0.2).(Landis and Koch 1977) The physical features measured in selection of variables for later use in For purposes of comparison, this manner were derived from the operational definitions, and Cronbach’s alpha is also reported for urban design literature, from earlier ultimately, for the selection urban these ratings. visual assessment studies, and most design qualities to be operationalized. importantly, from interviews with our To assess inter-rater reliability of 10. Analyze Content of national expert panel. As panelists measured physical features, a Sampled Scenes were rating scenes, they were also random sample of video clips was Our objective from the beginning of commenting on the physical features assigned to three other members of this project was to operationalize that caused ratings to be high or low the research team. The sample urban design qualities in objective, with respect to each urban design consisted of 12 clips in all, or four per quantitative terms. This meant that quality. Interviews, which had been team member. Sample size was the ratings by the expert panel, taped, were reviewed to identify limited by the time required to promising variables. evaluate more than 100 features of insofar as possible, had to be each clip. explained in terms of measurable

15 Before they evaluated the clips in was almost perfect agreement (ICCs Table 4 had insufficient variance their samples, the other team > 0.8) or substantial agreement (0.8 across the sample to compute inter- members discussed the measurement > ICCs > 0.6) among the team rater reliability statistics. of each feature with the first two members. It is relatively easy to team members via conference call. count objects and measure widths. A major task in the second phase of As these other team members Several features had low or even this study will be to refine operational evaluated each clip in their sample, negative ICC values. Of these, definitions and provide photographic they continually referred back to the features such as the number of examples of physical features that operational definitions, a process that landscape elements could probably proved to be powerful predictors of added considerably to the time be rated more consistently with urban design quality ratings but were required (particularly for the PI from better operational definitions. Other not rated reliably in the first round. public health, who had never done features, such as landscape anything quite like it). condition, involve a high degree of judgment and might require training ICC and Chronbach’s alpha values for and/or photographic examples to physical features are presented in achieve reasonable inter-rater Table 4. For most features, there reliability. Those missing values in

Table 4: Inter-rater Reliability for Estimates of Physical Features*

Variable ICC Alpha Variable ICC Alpha Variable ICC Alpha proportion street number of number of small .471 .611 wall – opposite .588 .737 .968 .982 courtyards etc. planters side number of landscape arcades ------.389 .640 -.115 .244 enclosed sides condition average building common tree number of .763 .878 setback – same .215 .338 spacing – same .766 .867 landmarks side side number of common tree major common building ------.814 .897 spacing – both .283 .407 landscape setbacks sides features

16 Variable ICC Alpha Variable ICC Alpha Variable ICC Alpha number of memorable building height – ------.741 .864 moving .895 .946 architecture same side pedestrians distinctive building height number of 1.000 1.000 .855 .940 .728 .865 signage to width ratio people standing number of long building height – number of .585 .714 .939 .966 .994 .997 sight lines opposite side people seated

common building terminated vista .436 .571 .500 .646 noise level .571 .704 heights proportion common progress toward .833 .906 .690 .833 outdoor dining 1.000 1.000 buildings masses intersection proportion progress toward ..718 .860 street width .870 .927 number of tables .916 .954 distant point number of street -.110 -.296 median width .007 .014 number of seats -.052 -.065 connections number of number of .913 .951 sidewalk width .693 .807 pedestrian street .938 .967 buildings lights building height number of other number of land .762 .852 to street width .894 .947 pieces of street .933 .969 uses ratio furniture proportion sidewalk clear number of misc historic .518 .720 .506 .690 .849 .940 width street items buildings number of number of pieces buildings with .876 .934 buffer width .542 .772 .529 .748 of public art ID

17 Variable ICC Alpha Variable ICC Alpha Variable ICC Alpha proportion of number of number of traffic buildings with .443 .721 .641 .861 .876 .946 paving materials signs ID various building textured number of place .845 .916 1.000 1.000 .761 .865 ages sidewalk surface or business signs number of textured street number of building .409 .550 1.000 1.000 .766 .867 surface directional signs materials number of pavement number of .503 .642 .627 .771 1.000 1.000 building colors condition billboards number of .350 .609 debris condition .532 .681 common signage ------accent colors number of number of building .610 .775 .958 .984 graffiti ------parked cars projections number of proportion proportion sky .891 .937 .965 .991 .831 .899 visible doors parked cars ahead number of number of proportion .600 .818 .970 .984 .550 .690 recessed doors moving cars buildings ahead proportion of average speed of proportion .531 .790 .862 .919 .602 .732 recessed doors moving cars pavement ahead proportion first number of proportion cars floor facades .841 .911 ------.174 .338 moving cyclists ahead with windows proportion number of curb proportion street overall façades .643 .764 .814 .897 .837 .906 extensions furniture ahead with windows

18 Variable ICC Alpha Variable ICC Alpha Variable ICC Alpha common number of proportion window .845 .916 midblock ------landscaping .911 .949 proportions crossings ahead number of number of proportion sky awnings or .717 .828 midblock pass------.943 .976 across overhangs throughs proportion of proportion building height -.122 -.064 overhead utilities ------.887 .935 buildings across interruptions number of non- number of proportion rectangular .399 .738 landscape -.086 -.019 .700 .808 pavement across silhouettes elements proportion non- landscaped proportion cars rectangular -.040 .251 1.000 1.000 .939 .967 median across silhouettes common proportion street architectural .431 .675 number of trees .804 .883 .633 .750 furniture across styles proportion common number of tree .585 .714 .649 .854 landscaping .894 .946 materials wells across proportion of proportion of .795 .878 .922 .956 active uses shaded sidewalk proportion number of large street wall – .938 .976 .349 .510 planters same side *For more complete variable labels and operational definitions, see Appendix 3.

19 features. This may not be the best particular scene may have evoked a 11. Relate Urban Design approach. particularly positive or negative Ratings to Measurable reaction in a particular viewer. We Physical Features When an outcome varies viewed such unique reactions as systematically in two dimensions, and measurement errors. Multivariate statistical methods were random effects are present, the used to model urban design ratings in resulting data structure is best The more interesting source of terms of measurable physical represented by a cross-classified variation in scores is that associated features of scenes. Models were random effects model. For an with scenes. Indeed, the purpose of specified based on hypothesized introduction to this class of models, this study is to identify the physical relationships between urban design readers are referred to Chapter 12 in features of scenes that give rise to qualities and specific physical Raudenbush and Bryk's Hierarchical high or low ratings on urban design features. These in turn were partly a Linear Models: Applications and Data quality scales. In statistical parlance, matter of common sense, partly a Analysis Methods. the "scene effect" gives rise to "scene reflection of the urban design variance." While not of much literature, and partly a product of the The outcome variable in this analysis interest, variation also occurs across interviews with the expert panelists. was the urban design quality rating panelists and must be accounted for. To keep model building from assigned by an individual panelist to Again in statistical parlance, the becoming a data mining exercise, a an individual street scene. Where all "viewer effect" gives rise to "viewer matrix of hypothesized relationships 48 scenes were rated by all 10 variance." The unique reactions of was created and only the features panelists, our sample consisted of individual panelists, and the random plausibly linked to urban design 480 ratings. For one perceptual variations in their scoring across qualities were actually tested for quality, linkage, one panelist declined scenes, produce "measurement error predictive power. Appendix 4 to provide ratings for all video clips, variance." contains the matrix. Xs signify and the sample was slightly smaller. plausible links. Xs are bolded if In order to bring into focus the physical features proved significantly Ratings varied from scene to scene interesting variation, that is the related to urban design qualities. due to different qualities of the street variation across street scenes, it Cross-Classified Random Effects itself and its edge. Ratings also helps statistically to separate the Models varied from viewer to viewer due to scene variance from viewer variance differences in judgment. Some and measurement error variance. Visual assessment studies often viewers were more generous in their Doing so, we are able to eliminate employ multiple regression analysis grading than others. Finally, ratings viewer effects when evaluating the to explain scene ratings in terms of varied due to unique interactions power of physical features to predict objectively measured physical between scenes and viewers. A street scene ratings. If we had simply

20 Table 5: Variance in Ratings by Source for Each Urban Design Quality used the average ratings of scenes (% of total variance in parentheses) as the outcome variable, and the Scene Viewer Measurement physical features of scenes as Variance Variance Error Total Variance explanatory variables, the effect of 0.67 0.16 0.50 1.33 scene variance might have been imageability confounded by the effect of viewer (50) (12) (38) 0.46 0.17 0.55 1.18 variance. legibility (39) (14) (47) 0.83 0.10 0.48 1.41 Our analysis began by partitioning enclosure the total variance in urban design (59) (7) (34) 0.68 0.11 0.50 1.29 quality ratings among the three human scale sources of variation—scenes, viewers, (53) (8) (39) 0.77 0.13 0.62 1.52 and measurement errors. The model transparency consisted of two parts: (51) (8) (41) 0.51 0.26 0.74 1.51 linkage actual rating = predicted (34) (17) (39) 0.6 0.09 0.47 1.16 rating + measurement error complexity (52) (8) (40) 0.45 0.11 0.62 1.18 where the actual rating is the sum of coherence the predicted score for a given scene (38) (9) (53) 0.46 0.17 0.43 1.06 by a given viewer plus the tidiness measurement error; and (43) (16) (41) shows the total variance in ratings 0.50. The total variance was thus predicted rating = constant + and the portions attributable to each split in the following proportions: viewer effect + scene effect source. The fuzzier constructs such 50 percent scene variance, as legibility and linkage have higher 12 percent viewer variance, and where the predicted rating is just the proportions attributable to viewer 38 percent measurement error sum of a constant plus a viewer judgment and measurement error. variance. effect and a scene effect. As an example, for the perceptual For all perceptual qualities, there was These equations were estimated quality of imageability, the scene more variance across scenes than using HLM 5.90 software, a statistical variance was 0.67, the viewer across viewers. This is not unusual in package developed by Raudenbush, variance was 0.16, and the visual assessment surveys. Bryk, Cheong, and Congdon (2000). measurement error variance was A set of additional models was For each perceptual quality, Table 5 estimated in order to reduce the

21 unexplained variance in perceptual only available variables characterizing proportions. See Appendix 3 for quality ratings. These models viewers—urban designer or not (1 or variable definitions. included characteristics of viewers 0 dummy) and new urbanist or not (1 and scenes: or 0 dummy)—proved to have no In all, 37 physical features proved explanatory power in most analyses. significant in one or more models. actual score = predicted score That is to say, neither variable was Six features were significant in two + measurement error significant at the 0.10 probability models: long sight lines, number of level, except in the model for human buildings with identifiers, proportion exactly as above; and scale, in which the variable for urban first floor façade with windows, designer proved marginally proportion active uses, proportion predicted score = constant + significant. Apparently urban street wall–same side, and number of viewer random effect + scene designers and others, and new pieces of public art. Two features random effect + a*viewer v urbanists (a subset of the designers) were significant in three models: ariables + b*scene variables and others, react similarly to street number of moving pedestrians and scenes. This is consistent with earlier presence of outdoor dining. The where the viewer random effect is the visual assessment literature revealing models for each quality are presented portion of the viewer effect left common environmental preferences and discussed below. unexplained by viewer across professions. Discussion characteristics, the scene random effect is the portion of the scene By contrast, many of the variables Imageability effect left unexplained by scene characterizing scenes proved For imageability, the estimated model characteristics, viewer variables is significant individually and in left the measurement error variance the vector of relevant viewer combination with each other. This unchanged at 0.50, reduced the characteristics, a is the vector of again is consistent with the visual unexplained viewer variance only associated coefficients, scene assessment literature. The models slightly from 0.16 to 0.15, but variables is the vector of relevant that reduced the unexplained reduced the unexplained scene scene characteristics, and b is the variance of scores to the greatest variance substantially, from 0.67 to vector of associated coefficients. degree, and for which all variables 0.19. Altogether, 72 percent of the These variables capture the "fixed had the expected signs and were variation across scenes, and 37 effects" of viewers and scenes on significant at the 0.10 level or percent of the overall variation in urban design ratings. beyond, are presented in Tables 6 imageability scores (including through 14 . Most of the Results of Statistical Analysis variation across viewers and independent variables in these tables measurement errors), were explained Many combinations of viewer and are object counts, though there are by the significant scene variables scene variables were tested. The also dummy variables and (Table 6). All of the significant scene

22 variables had acceptable levels of Table 6: Best-Fit Imageability Model inter-rater reliability (with intraclass Variable Coefficient t- p- correlation coefficients of 0.40 or statistic value above, except for major landscape features, which had insufficient constant 2.516 variance across the sample to courtyards/plazas/parks (#) 0.393 3.58 0.001 compute inter-rater reliability). The major landscape features (#) 0.735 2.00 0.046 significance of the number of proportion of historic buildings 0.948 4.16 0.000 pedestrians and outdoor dining points buildings with identifiers (#) 0.115 1.80 0.072 to the importance of human activity buildings with non-rectangular silhouettes (#) 0.0745 1.95 0.052 in creating imageable places. The lack of significance of landmarks, pedestrians (#) 0.0271 4.73 0.000 memorable architecture, and public noise level (rating) -0.195 -2.11 0.035 art forces us to rethink just what outdoor dining (y/n) 0.703 3.97 0.000 makes a place memorable. Overall, Proportion of Scene Variance Explained 0.72 the model is strong. Proportion of Total Variance Explained 0.37 Legibility Table 7: Best-Fit Legibility Model For legibility, the estimated model Variable Coefficient t- p-value left the measurement error variance statistic and unexplained viewer variance unchanged (Table 5). It reduced the constant 2.412 unexplained scene variance from memorable architecture (y/n) 0.620 2.49 0.013 0.46 to 0.21, accounting for only 54 terminated vista (y/n) 0.722 3.57 0.001 percent of the scene variance and 21 buildings with identifiers (#) 0.228 3.55 0.001 percent of the total variance, the common tree spacing and type – same side 0.433 2.68 0.008 lowest percentages among the nine (y/n) urban design qualities studied (Table public art (#) 0.342 2.07 0.039 7). All of the significant scene place/building/business signs (#) 0.0537 2.18 0.030 variables had acceptable levels of inter-rater reliability, except for Proportion of Scene Variance Explained 0.54 memorable architecture, which had Proportion of Total Variance Explained 0.21 insufficient variance across the sample to compute inter-rater reliability. The number of buildings

23 Table 8: Best-Fit Enclosure Model imageability, suggesting that panelists may have had difficulty Variable Coefficient t- p- distinguishing between these two statistic value concepts. As noted earlier, legibility constant 2.570 itself had a low level of inter-rater long sight lines (#) -0.308 -2.12 0.035 reliability. Overall, the model is proportion street wall – same side 0.716 3.51 0.001 weak. proportion street wall – opposite side 0.940 3.17 0.002 Enclosure proportion sky ahead -1.418 -1.92 0.055 For enclosure, the estimated model proportion sky across -2.193 -2.32 0.021 left the measurement error variance Proportion of Scene Variance Explained 0.72 unchanged, reduced the unexplained Proportion of Total Variance Explained 0.43 viewer variance slightly from 0.10 to 0.09, and reduced the unexplained Table 9: Best-Fit Human Scale Model scene variance from 0.83 to 0.23. This is the largest absolute reduction Variable Coefficient t- p- in unexplained scene variance. With statistic value just five variables, the model for constant 2.612 enclosure explains 72 percent of the urban designer (y/n) 0.382 1.84 0.066 scene variance and 43 percent of the long sight lines (#) -0.775 -4.97 0.000 total variance (Table 8). All of the proportion first floor with windows 0.916 2.93 0.004 significant variables have high levels proportion active uses 0.306 1.77 0.077 of inter-rater reliability, with ICCs building height – same side -0.00308 -2.08 0.038 above 0.59. The signs of the coefficients in the model are as small planters (#) 0.0469 1.86 0.063 expected, with long sight lines, miscellaneous street items (#) 0.0635 3.25 0.002 proportion of the view ahead that is Proportion of Scene Variance Explained 0.62 sky, and proportion of the view Proportion of Total Variance Explained 0.35 across the street that is sky detracting from the perception of enclosure. A more continuous “street with identifiers and the number of easily explained but may be related wall” of building facades, on each signs have obvious conceptual to the ability to place the street in a side of the street, adds to the connections to legibility; the larger spatial context. The set of perception of enclosure. This model significance of common tree spacing variables in the model also has suggests that enclosure is influenced and memorable architecture is less conceptual connections to

24 not just by the near side of the street Table 10: Best-Fit Transparency Model but also by views ahead and across Variable Coefficient t- p- the street. Surprisingly, the average statistic value street width, average building setback, average building height, and constant 1.709 relationship between the width of the proportion first floor with windows 1.219 3.13 0.002 street and building height were not proportion active uses 0.533 2.96 0.004 significant. Overall, the model is proportion street wall – same side 0.666 2.57 0.011 strong. Proportion of Scene Variance Explained 0.62 Human Scale Proportion of Total Variance Explained 0.32 For human scale, the estimated model left the measurement error Table 11: Best-Fit Linkage Model variance unchanged, reduced the Variable Coefficient t- p- unexplained viewer variance from statistic value 0.11 to 0.08, and reduced the constant 2.104 unexplained scene variance from street connections to elsewhere (#) 0.623 3.31 0.001 0.68 to 0.26. Seven variables explain visible doors (#) 0.134 3.36 0.001 62 percent of the scene variance and 35 percent of the total variance in proportion recessed doors 0.613 3.11 0.002 human scale (Table 9). All of the common building heights (y/n) 0.576 3.52 0.001 significant variables have ICCs of outdoor dining (y/n) 0.415 2.21 0.027 0.59 or higher. The signs of the Proportion of Scene Variance Explained 0.61 coefficients are as expected: the Proportion of Total Variance Explained 0.21 number long sight lines and building height on the same side of the street decrease the perception of human viewers are significant: if the viewer variance and unexplained viewer scale, while the presence of first floor is an urban designer, the rating of unchanged, and reduced the windows, small planters, and street human scale is higher, all else equal. unexplained scene variance from items increase the perception of Overall, the model is strong. 0.77 to 0.29. Just three variables explain 62 percent of the scene human scale. Human activities are Transparency also important, specifically the variance and 32 percent of the total proportion of street frontage with For transparency, the estimated variance in transparency: the active uses. Human scale is the only model left the measurement error proportion of the first floor with quality for which characteristics of windows, the proportion of active

25 uses, and the proportion of street Table 12: Best-Fit Complexity Model wall on the same side (Table 10). All Variable Coefficient t- p- three variables have acceptable levels statistic value of inter-rater reliability. The model suggests that being able to see into constant 1.453 buildings and having human activity buildings (#) 0.0458 2.42 0.016 along the street frontage both dominant building colors (#) 0.225 2.74 0.007 contribute to the perception of accent colors (#) 0.115 2.21 0.027 transparency. Note that windows pedestrians (#) 0.0311 5.96 0.000 above ground-level do not increase outdoor dining (y/n) 0.418 2.30 0.022 the perception of transparency (after controlling for other variables). public art (#) 0.286 1.96 0.051 Overall, the model is strong. Proportion of Scene Variance Explained 0.73 Linkage Proportion of Total Variance Explained 0.38 For linkage, the estimated model left significance of recessed doors, levels of inter-rater reliability, and the measurement error variance and outdoor dining, and common building the signs on the coefficients are in unexplained viewer unchanged, and heights on opposite sides of the the expected direction. The reduced the unexplained scene street suggests the importance of significance of pedestrians and variance from 0.51 to 0.20. The psychological as well as physical outdoor dining suggests that human model for linkage, with five variables, connections between buildings, activity may contribute as much to explains 61 percent of scene variance sidewalks, and streets. Overall, the the perception of complexity as do but only 21 percent of total variance model is weak. physical elements. The lack of (Table 11). Linkage has the highest significance of several other variables Complexity view variance and measurement is notable: number of building error of the nine urban design For complexity, the estimated model materials, number of building features, and is tied with legibility for left the measurement error variance projections, textured sidewalk the smallest percentage of total and viewer variance unchanged, surfaces, number of streets lights and variance explained. These statistics while reducing unexplained scene other kinds of street furniture, among indicate lack of clarity in the concept variance from 0.67 to 0.19. Six others. Overall, the model is strong. of linkage. Four of the five variables variables explain 73 percent of scene Coherence in the model had acceptable levels of variance and 38 percent of total inter-rater reliability; the number of variance for complexity (Table 12). For coherence, the estimated model street connections to other places Except for the number of accent left the measurement error variance was the notable exception. The colors, all variables have acceptable and unexplained viewer variance

26 unchanged, and reduced the Table 13: Best-Fit Coherence Model unexplained scene variance from Variable Coefficient t- p- 0.45 to 0.15. Only four variables statistic value were significant in the model for coherence (Table 13). These Constant 2.495 variables explained 67 percent of the common window proportions (y/n) 0.979 6.18 0.000 scene variance but only 25 percent of common tree spacing and type – both 0.356 2.41 0.016 the total variance. All variables sides (y/n) except common tree spacing have pedestrians (#) 0.0217 4.29 0.000 ICCs over 0.85, indicating a high pedestrian scale street lights (#) 0.0566 1.81 0.070 degree of inter-rater reliability. Two Proportion of Scene Variance Explained 0.67 of the variables have strong Proportion of Total Variance Explained 0.25 conceptual connections to coherence: common window proportions and Table 14: Best-Fit Tidiness Model common tree spacing and type on both sides of the street. Connections Variable Coefficient t- p- to the other two variables are less statistic value obvious. Pedestrian scale street pavement condition (rating) 0.197 3.31 0.001 lights are always of uniform style and debris condition (rating) 0.272 3.84 0.000 size and unify scenes visually to a overhead utilities (y/n) -0.638 -2.34 0.020 surprising degree. Pedestrians landscape condition (rating) 0.230 4.29 0.000 become a dominant and relatively Proportion of Scene Variance Explained 0.70 uniform element as their numbers increase. Other conceptually Proportion of Total Variance Explained 0.30 important variables are missing from the model, including common unexplained scene variance from variability in overhead utilities was architectural styles and common 0.46 to 0.14. The model for tidiness not large enough to compute inter- building masses. Overall, this model explained 70 percent of scene rater reliability. The coefficients of all is weak. variance and 30 percent of total explanatory variables have the variance with just four variables expected signs, and the variables are Tidiness (Table 14). Two of these variables, those with the strongest conceptual For tidiness, the estimated model left ratings of pavement condition and connections to tidiness. Overall, the the measurement error variance and debris condition, had acceptable model is strong, although inter-rater unexplained viewer variance inter-rater reliability; the rating of reliability for landscape condition is a unchanged, while reducing the landscape condition did not, and the concern.

27 12. Select Which Qualities Table 15: Performance of Urban Design Qualities Relative to Selection Criteria to Define Operationally Portion of Scene Inter-rater In Section 8, we established criteria Variance/Total Reliability of for deciding which urban design Relationship to Inter- Variance Significant qualities to operationalize. They Walkability in rater Explained by Variables Best-Fit Model reliability Best-Fit (number with Criteria were: significant relationship of the (p-value) (ICC) Models ICC>0.4) Met quality to overall walkability (p < Imageability 0.000 0.494 0.72/0.37 7 of 7 5 of 5 0.05); ability to measure the quality (1 missing) with at least moderate inter-rater reliability (ICC > 0.4); ability to Legibility --- 0.380 0.54/0.21 5 of 5 1 of 5 explain overall variations in the (1 missing) quality to a moderate degree with Enclosure 0.016 0.584 0.72/0.43 5 of 5 5 of 5 measurable scene variables (explained portion > 0.3); ability to Human scale 0.000 0.508 0.62/0.35 7 of 7 5 of 5 explain scene-specific variations in the quality to a substantial degree with measurable scene variables Transparency 0.023 0.499 0.62/0.32 3 of 3 5 of 5 (explained portion > 0.6); and ability to measure these same variables with Linkage --- 0.344 0.61/0.21 4 of 5 1 of 5 at least moderate inter-rater reliability (ICC > 0.4). A Complexity --- 0.508 0.73/0.38 5 of 6 3 of 5 performance summary for each quality with respect to each criterion is presented in Table 15. Coherence --- 0.374 0.67/0.25 3 of 4 1 of 5

According to these criteria, the Tidiness 0.117 0.421 0.70/0.30 2 of 3 3 of 5 qualities of imageability, enclosure, (1 missing) human scale, and transparency have great potential for operationalization. given no further consideration. The has thus decided to include these two They meet all five criteria. The qualities of complexity and tidiness urban design qualities, along with the qualities of legibility, linkage, and fall somewhere between the first four qualities that meet all coherence have very little potential extremes, meeting three of five criteria, in the field survey for operationalization, each meeting criteria; tidiness comes close to instrument. only one of five criteria. They will be meeting a fourth. The research team

28 Figure 6: Introduction to One Urban Design Quality, Imageability Figure 7: Instructions for Measuring a Physical Feature Related to Imageability The instrument showed users how to quality was provided. This definition 13. Develop a Draft measure physical features related to was based on the urban design Instrument for Measuring each of these qualities and how to literature and was refined with the Urban Design Qualities convert these measurements into help of the expert panel of urban urban design quality scores based on designers and top professionals from A draft measurement instrument was the statistical models described in related fields. prepared. It focused on the six urban Section 11. Each urban design design qualities that met three or quality in the manual was presented (b) Expert panel comments: Our more of the performance criteria with a set of instructions (see Figure research on urban design qualities outlined in Table 15. Again, the 6). All of the instructions followed the included extensive interviews with qualities were: same format: and surveys of the expert panel. We reported a sample of what they had imageability The first page of instructions to say about each quality. enclosure introduced users to the urban design human scale quality with: (c) Photographic examples: Two transparency contrasting photos were shown to complexity (a) Qualitative definition: A short and illustrate extreme examples of each tidiness concise definition of the urban design urban design quality, as judged by

29 the expert panel. Short descriptions were also provided pointing out the features that made each scene either high or low with respect to the quality.

The next few pages of instructions showed users how to measure the urban design quality. Measuring urban design qualities involves visiting streets and being able to identify and count certain street features (see Figure 7). Users also need to make educated estimates of other features. Detailed illustrations were provided for each measurement Figure 8: Portion of the Urban Design Qualities Scoring Sheet needed to arrive at a value of the urban design quality. field with graduate students to test urban design qualities. We did this measurement protocols. Scenes that for a sample of 16 street scenes from The last page of the manual provided were part of the original visual our original set of 48 scenes. a scoring sheet that can be used on assessment survey were used to Resulting field measurements were the field to record measurements and assess whether measurements of compared to our “gold standard” calculate urban design quality scores physical features in the field were estimates based on video clips. Field for the street segment in question. consistent with measurements in the observations and video clips were The scoring sheet summarized field lab using video clips. We were compared for: (1) inter-rater measurements and provided a attempting to validate the use of reliability of individual multiplier to be applied to each video clips in our earlier urban design measurements; (2) inter-rater relevant physical feature in order to quality analyses. We were also reliability of urban design quality compute an overall urban design seeing how the protocol used to scores based on the individual quality score (see Figure 8). shoot clips would translate into a measurements: (3) rank-order procedure for field measurements. correlations of individual 14. Test and Refine the measurements (assuming that Draft Instrument In the field, we measured all physical With the draft instrument available, features that proved significant principal investigators went into the contributors to the six remaining

30 Table 16: Initial Field Manual Test Results relative ranking of scenes might be comparable even if absolute values Spearman differ between field observations Alpha ICC r p Imageability 0.682 0.111 0.49 0.055 and video clips); and (4) rank-order 1. number of courtyards, plazas, and parks (both sides, within study 0.747 0.557 0.58 0.018 correlations of urban design quality 2. number of major landscape features (both sides, beyond study area) not enough variance scores (assuming again that 3. proportion historic building frontage (both sides, within study area) 0.808 0.558 0.68 0.004 relative rankings might be 4. number of buildings with identifiers (both sides, within study area) 0.13 -0.404 0.24 0.379 comparable even if absolute values 5. number of buildings with non-rectangular shapes (both sides, within 0.876 0.775 0.81 0.000 differ). Table 16 summarizes our 6. presence of outdoor dining (your side, within study area) 0.921 0.854 0.86 0.000 7. number of pedestrians (your side, within study area) 0.782 0.646 0.51 0.044 results. 8. noise level (both sides, within study area) 0.454 0.284 0.3 0.263 Enclosure 0.73 0.175 0.41 0.111 What we found were major 1. number of long sight lines (both sides, beyond study area) 0.457 0.129 0.34 0.194 discrepancies between 2a. proportion street wall (your side, beyond study area) 0.837 0.731 0.83 0.000 measurements in the field and the 2b. proportion street wall (opposite side, beyond study area) 0.466 0.298 0.51 0.042 lab for certain physical features, 3a. proportion sky (ahead, beyond study area) 0.634 0.361 0.56 0.026 3b. proportion sky (across, beyond study area) 0.605 0.197 0.33 0.206 and hence significant discrepancies Human Scale 0.662 0.515 0.78 0.000 for the urban design qualities to 1. number of long sight lines (both sides, beyond study area) 0.457 0.129 0.34 0.194 which they contribute in our scoring 2. proportion windows at street level (your side, within study area) 0.818 0.559 0.69 0.003 formulas. Discrepancies were 3. proportion active uses (your side, within study area) 0.797 0.678 0.64 0.007 significant for the following qualities 4. average building heights (your side, within study area) 0.864 0.767 0.77 0.000 and contributing features (the latter 5. number of small planters (your side, within study area) 0.796 0.594 0.61 0.012 6. number of miscellaneous street items (your side, within study area) 0.489 0.22 0.64 0.008 in parentheses): Transparency 0.891 0.785 0.9 0.000 ƒ imageability (number of 1. proportion windows at street level (your side, within study area) 0.82 0.568 0.68 0.004 buildings with identifiers and 2. proportion street wall (your side, beyond study area) 0.833 0.727 0.82 0.000 noise level) 3. proportion active uses (your side, within study area) 0.797 0.678 0.64 0.007 ƒ enclosure (number of long sight Complexity 0.57 -0.199 0.48 0.06 lines and proportion sky across 1. number of buildings (both sides, beyond study area) 0.869 0.342 0.7 0.003 2a. number of primary building colors (both sides, beyond study area) 0.143 -0.315 0.07 0.811 the street) 2b. number of accent colors (both sides, beyond study area) 0.413 -0.02 0.41 0.117 ƒ human scale (number of long 3. presence of outdoor dining (your side, within study area) 0.842 0.692 0.77 0.000 sight lines) 4. number of pieces of public art (both sdies, within study area) not enough variance ƒ complexity (number of primary 5. number of pedestrians (your side, within study area) 0.858 0.675 0.83 0.000 building colors and number of Tidiness 0.058 -0.167 -0.01 0.959 accent colors) 1. pavement condition (your side, within study area) 0.735 0.215 0.59 0.017 2. debris condition (your side, within study area) -0.203 -0.105 -0.08 0.758 ƒ tidiness (debris condition and 3. overhead utilities (both sides, within study area) not enough variance landscape condition) 4. landscape condition (your side, within study area) 0.391 0.114 0.28 0.293

31 The time between the filming of video across the street. We could not clips and the field validation distinguish different shades of colors. accounted for some of the We could not see buildings in the discrepancies. More than a year had distance. passed, and validations often occurred at a different time of day, To deal with discrepancies, we day of the week, and season of the had four options: (1) ignore them on year. Figure 9 compares one scene the assumption that the relationships at the time of original filming to the between urban design qualities and same scene at the time of field physical features are as estimated validation. Occurring after a long from the clips, even if measurements winter, the validation process found differ; (2) refine field measurement (a) many streets stark, de-populated, protocols to more close approximate and in need of maintenance. measurements based on clips; (3) drop physical features that could Other discrepancies arose from the not be measured consistently in the greater distance observers could field, and re-estimate the formulas travel in about the same time when without these features; or (4) drop simply walking rather than shooting urban design qualities that could not video clips. Count totals tended to be be estimated consistently due to higher in the field than the lab inconsistent measures of because the field survey protocol component physical features. took observers farther down the (b) block. For some scenes, counts were Option (2) was preferred where much higher. For others, they were feasible. For certain features, only marginally so. changes in measurement protocols were implemented through changes Still other discrepancies were in the field survey manual. Number inherent in the medium used in the of building colors, for example, could lab, that is, in the video clips be estimated more consistently by themselves. Shadows, glare, and instructing users to count only the panning limited what could be seen in number of basic colors, not shadings. the clips, particularly on the opposite Noise levels could be estimated more side of the street and ahead in the consistently by instructing users to distance. We could not read signs (c) Figure 9: Comparison of One Scene at Time of Original Filming (a and b) and at Time of Field Validation (c) 32 average noise levels over several The combined variable “all street dropping them from the tidiness passes of study area. furniture and other street items” was formula, would have left only two substituted for “miscellaneous street significant physical features upon For two features, field experience items” in the model of human scale. which to base tidiness scores. Given taught us to combine elements This variable is the sum of number of the weak relationship between treated separately in our original tables, number of seats, number of tidiness and walkability ratings by our analyses. The distinction between pedestrian-scale street lights, expert panel, we decided to drop people walking, standing, and sitting number of pieces of other street tidiness from the final version of the struck us artificial as we walked the furniture, and number of field survey manual. same stretch of street several times. miscellaneous street items. It Walkers became standers etc. The improved the overall explanatory 15. Train Lay Observers in distinction among the different power of the human scale model, and the Classroom categories of street furniture and caused one of the variables that had With the final field survey instrument miscellaneous street items was been significant—proportion of active in hand, lay observers were given difficult to keep straight. Parking uses along the street—to no longer classroom training in its use. The meters and trash cans were in one be so. The proportion of active uses classroom training took three hours. category, hydrants and ATMs in is highly correlated with the new The lay observers were students from another. Tables, seating, and street combined “all street furniture and UC Davis. Eight students participated lights were in a third, fourth, and fifth other street items” variable. The new in the training. A sub-sample of categories. combined variable was also tested in video clips from the original visual the complexity model and proved assessment survey were used as the So we decided to combine categories insignificant. training medium in the classroom, and test the resulting variables in our where students could compare their urban design quality models. The For one urban design quality— measurements of physical features combined variable “people” within a tidiness—field and lab estimates were with the gold standard established by scene was substituted for “moving inconsistent even as measured by the the research team. pedestrians” in the models of Spearman rank-order correlation imageability and complexity. It had a coefficient. Two of the component The protocol for the classroom slightly higher significance level in physical features, debris condition training was as follows: the imageability model, without and landscape condition, contributed greatly affecting the relationship of to the low inter-rater reliability. Items needed for classroom training: other variables to imageability These features are highly variable ƒ laptop for trainer ratings. It did not perform as well as over time and more subjective than ƒ digital projector moving pedestrians in the complexity most others in our urban design ƒ 32-clip DVD (comes with training model. formulas. Exercising option (3), and package)

33 ƒ scoring sheets with “gold hard and fast, only as values arrived 16. Field Test the Survey standard” measurements for test at through a careful process. Instrument with Lay clips (come with training package) ƒ field survey manuals for Step 3. Make Independent Observers participants Measurements for Additional Test After the UC Davis students were Clips trained in the classroom, they were Step 1. Review Field Survey Manual For another video clip (clip #27), sent to the field to complete Trainees were given copies of the trainees made measurements on observations for selected street field survey manual and were their own, filling out a blank scoring segments. Ten segments in introduced to its contents. For each sheet as they went along. The clip downtown Davis, CA and six urban design quality, the component was replayed as many times as segments in downtown Sacramento, physical features were reviewed, and required for them to complete the CA were used for this field test. the measurement protocols were task. When all trainees were done, These segments were chosen to described. Trainees were shown the trainer read, and trainees achieve as much variation as possible scoring sheets at the end of the recorded, the gold standard in the measured qualities. Two manual. They were encouraged to measurements next to trainees’ own students were assigned to each ask questions. measurements. The clip was then segment to enable an analysis of reviewed to reconcile differences in inter-rater reliability. A total of 32 Step 2. Review Gold Standard measurements. Again, gold standard observations were thus completed. Measurements for Test Clips measurements were not presented as Raters were debriefed after A video clip (clip #14 on the DVD) indisputable. Much time was spent completing their observations and from the original expert panel visual discussing differences. This process provided suggestions for clarifying assessment survey was shown to could be repeated with additional instructions in the field manual. trainees via DVD and digital clips (“gold standard scores” have projector. The first scoring sheet is been provided for clip #’s 31, 25, 16, ICC and Chronbach’s alpha values for filled out with the research team’s 12, 24, and 29) until all trainees, in a physical features and for urban “gold standard” measurements. For show of hands, express confidence in design quality scores are presented in each physical feature on the scoring their ability to measure physical Table 17. For half of the features, sheet, the gold standard features of scenes consistently. there was almost perfect agreement measurement was reviewed as the (ICCs > 0.8) or substantial clip was played. The clip was agreement (0.8 > ICCs > 0.6) replayed as many times as required between the raters. For four more to review all physical features on the features, there was good agreement scoring sheet. The gold standard (0.6 > ICCs > 0.4). measurements were not presented as

34

Table 17: Field Test Results For seven features, agreement was Alpha ICC fair or poor for a variety of possible Imageability .927 .863 1. number of courtyards, plazas, and parks (both sides, within study .845 .584 reasons: 2. number of major landscape features (both sides, beyond study area) n/a n/a 3. proportion historic building frontage (both sides, within study area) .864 .750 ƒ Long sight lines: There was very 4. number of buildings with identifiers (both sides, within study area) .875 .769 little variation in the 5. number of buildings with non-rectangular shapes (both sides, within .899 .818 measurements for this feature. 6. presence of outdoor dining (your side, within study area) .887 .809 Only four values are possible, and 7. number of pedestrians (your side, within study area) .960 .913 8. noise level (both sides, within study area) .618 .432 none of the test segments had Enclosure .232 .033 more than 2 long sight lines. 1. number of long sight lines (both sides, beyond study area) -.208 -.238 Still, the disagreement between 2a. proportion street wall (your side, beyond study area) .517 .373 raters was often large: on five 2b. proportion street wall (opposite side, beyond study area) .863 .725 segments, one rater indicated 0 3a. proportion sky (ahead, beyond study area) .330 .157 sight lines, while the other rater 3b. proportion sky (across, beyond study area) -.568 -.208 Human Scale .768 .491 indicated 2 sight lines. As a 1. number of long sight lines (both sides, beyond study area) -.208 -.238 result, we modified the 2. proportion windows at street level (your side, within study area) .798 .663 instructional language in the field 3. proportion active uses (your side, within study area) .422 .239 manual. 4. average building heights (your side, within study area) .956 .912 5. number of small planters (your side, within study area) .786 .622 ƒ Street wall: The poor results for 6. number of miscellaneous street items (your side, within study area) .547 .422 Transparency .817 .708 street wall were influenced by two 1. proportion windows at street level (your side, within study area) .798 .663 segments in particular. On one 2. proportion street wall (your side, beyond study area) .517 .373 segment, a five-story parking 3. proportion active uses (your side, within study area) .422 .239 garage abutted the sidewalk, Complexity .868 .780 although the ground floor was set 1. number of buildings (both sides, beyond study area) .592 .388 back several feet. On another 2a. number of primary building colors (both sides, beyond study area) .279 .188 2b. number of accent colors (both sides, beyond study area) .551 .331 segment, several detached 3. presence of outdoor dining (your side, within study area) .887 .809 buildings were set back several 4. number of pieces of public art (both sides, within study area) .677 .528 feet from the sidewalk. In both 5. number of pedestrians (your side, within study area) .836 .700 cases, the raters made different decisions about whether to count the buildings as a part of the street wall. With these two segments removed, the ICC for

35 this feature is 0.605. The field interpretation of which buildings raters, several refinements were manual was edited to more were visible enough to be made to the field manual. Although clearly specify which buildings counted. With the problematic the final version of the field manual contribute to the street wall. segment excluded from the was not retested for inter-rater analysis, the ICC increased to reliability, we believe that the ƒ Sky ahead and sky across: 0.685. refinements will improve reliability. Raters found these measurements In addition, a longer classroom ƒ Basic colors and accent colors: difficult to estimate in the field, training session that focuses on the The poor results for these although they were relatively easy problematic features should help to features can be partly explained to estimate from the video clips. increase inter-rater reliability. by the low variation. However, Differences between the raters the measurements for these stemmed from differences in the features were inconsistent for choice of exact location from most segments. Raters differed 17. Finalize the Instrument which they estimated proportion in their judgment of whether two sky and difficulty in judging the The instrument was finalized based colors were sufficiently different extent of their field of vision. The on the classroom and field to count as two colors; tans, field manual now suggests the experience. The field instrument, grays, and other neutrals seemed use of a cardboard frame to training materials, and final report particularly challenging. ensure consistency. are now available on-line at www.activelivingresearch.org. The ƒ Active uses: The poor results for field instrument includes: qualitative Among the urban design qualities, this feature are related to the definitions of urban design qualities; Imageability had the highest poor results for street walls. explanations and photographic reliability, closely followed by Because raters made different illustrations of physical features Complexity and Transparency. The judgments about which buildings relating to urban design qualities; reliability for Human Scale was lower fronted on the street, they came procedures for field observation and but still good. However, the up with different proportions of data collection; and scoring reliability for Enclosure was poor, buildings with active uses. When procedures for translating objectively given the poor reliability for sight the two problematic segments measured physical features into lines, sky ahead, and sky across. were excluded from the analysis, urban design quality scores. A DVD Improvements in reliability for these the ICC increased to 0.562. of video clips and sample scoring features would improve the reliability sheets are available as part of the ƒ Number of buildings: The results for Enclosure. training package. for this feature were fair. Large Based on these results and the differences for one segment can comments and suggestions of the be explained by differences in the

36 University. Mr. Ferrell is a Charter of CommunityViz™, the first GIS Appendix 1: Biosketches of Member of the Congress for the New based planning decision support Expert Panel Members Urbanism. His work is featured in the software to fully integrate virtual Victor Dover is a principal of Dover, book The New Urbanism by Peter reality with scenario design, impact Kohl & Partners, founded in 1987 and Katz. analysis and policy simulation. He based in South Miami, Florida. Mr. was made a Fellow of the American Dover earned his Bachelor of Mark Francis is Professor of Institute of Architects in 1990 for his Architecture degree from Virginia Landscape Architecture at the achievements in urban design and Tech and his Master’s Degree in Town University of California, Davis where performance zoning. & Suburb Design from the University he founded and directed the Center of Miami. He has been certified by for Design Research. Trained in Robert Lane is the Director of the the American Institute of Certified landscape architecture and urban Regional Design Program and the Planners and is a charter member of design at Berkeley and Harvard, his Healthy Communities Initiative at the the Congress for the New Urbanism. work is concerned with the design Regional Planning Association of New Mr. Dover and his partner Mr. Kohl and theory of urban places. He is York and New Jersey. Mr. Lane is the have been recognized by Architecture Associate Editor of the Journal of author of numerous urban design magazine as being “among the Architectural and Planning Research studies and town plans that country’s best architects and urban and serves on the editorial boards of emphasize compact mixed-use designers.” several journals including Landscape development, alternative forms of Journal, Environment and Behavior, mobility and other dimensions of Geoffrey Ferrell established his own Journal of Planning Literature and active living community design. urban design firm in 1992. Before Children and Youth Environments. Robert Lane is also the that, Mr. Ferrell worked as a His most recent books are Urban co-principal investigator on several Designer/Code Writer for Duany- Open Space and Village Homes "natural experiments" including Plater-Zyberk Architects and Town (Island Press 2003). measuring the impacts on activity Planners in Miami and as the Director levels of a new greenway in Stamford of Urban Design for the Treasure Michael Kwartler is the founding (Connecticut) and of new transit Coast Regional Planning Council in director of the Environmental services in New Jersey. Florida. Mr. Ferrell holds a Master of Simulation Center, a non-profit Architecture degree with a Certificate research laboratory created to Anne Vernez Moudon is Professor in American Urbanism from the develop innovative applications of of Architecture, Landscape University of Virginia, a Bachelor of information technology for Architecture, and Urban Design and Architecture from Oregon School of community planning, design and Planning at the University of Design, and a Bachelor of Science in decision-making. He conceived and Washington, Seattle. She is President Public Policy from Willamette directed the design and development of the International Seminar on

37 Urban Morphology (ISUF), a Faculty and specific Urban Design Plans all Department of Landscape Associate at the Lincoln Institute of over the United States. Mr. Nelessen Architecture and Environmental Land Policy, a Fellow of the Urban is a charter member of the Congress Planning at the University of Land Institute, a National Advisor to for the New Urbanism. His book California at Berkeley. Trained and the Robert Wood Johnson Visions for a New American Dream professionally registered in both city Foundation, and an active participant was published by the American planning and architecture, he in The Mayors’ Institute on City Planning Association. His current received the Ph.D. and M.C.P. Design. Dr. Moudon holds a B.Arch. book What People Want is in first degrees from the Massachusetts from the University of California, draft. Institute of Technology, and the Berkeley, and a Doctor ès Science Bachelor of Architecture and Bachelor from the Ecole Polytechnique John Peponis is an architect and a of Arts from the University of Fédérale of Lausanne, Switzerland. professor of architecture at the Minnesota. He is a Fellow of the Her published works include Built for Georgia Institute of Technology. He American Institute of Architects and a Change: Neighborhood Architecture received his Ph.D. in architecture at member of the American Institute of in San Francisco (MIT Press 1986), University College London in 1983. Certified Planners. His recent books Public Streets for Public Use He teaches regularly at the National include: Streets and the Shaping of (Columbia University Press (1991), Technical University of Athens, The Towns and Cities (with Eran Ben- and Monitoring Land Supply with University of London and the Joseph, 2003), City Sense and City Geographic Information Systems Chalmers University of Technology. Design (as editor and contributor with (with M. Hubner, John Wiley & Sons, He is a member of the editorial Tridib Banerjee) and Wasting Away 2000). boards of Environment and Planning (by Kevin Lynch). B: Planning and Design and The Anton Nelessen is founder and Journal of Architecture as well as a Daniel Stokols is Professor of president of the award winning firm, member of the steering and Planning, Policy, and Design and A. Nelessen Associates He served as refereeing committees for the bi- Dean Emeritus of the School of Social consultant on seven of the 10 Smart annual International Symposia on Ecology at the University of Growth awards given by the State of Space Syntax. Most recently he has California, Irvine. Dr. Stokols New Jersey. Mr. Nelessen has been a co-authored an overview of "space received his B.A. degree at the professor at Harvard University and syntax" for the Handbook of University of Chicago and his M.A. at the Rutgers University Department Environmental Psychology (Bechtel and Ph.D. degrees at the University of Urban Planning and Policy and Churchman eds). of North Carolina, Chapel Hill. He is Development since 1974. His past President of the Division of trademarked Visual Preference Michael Southworth is Professor in Population and Environmental Survey has been used to generate both the Department of City and Psychology of the American Comprehensive Plans, Master Plans Regional Planning and the Psychological Association (APA).

38 Dr. Stokols was a recipient of the Annual Educator Award from the International Facility Management Association in 1988, the Annual Career Award of the Environmental Design Research Association in 1991, and the UCI Lauds and Laurels Faculty Achievement Award in 2003.

39 does not necessarily denote a well-being results: the feeling that a Appendix 2: Qualitative grandiose civic structure or even a space is a thoroughly pleasant place Definitions of Urban Design large object. In the words of Lynch, it in which to be. Qualities can be “a doorknob or a dome.” What Imageability is influenced by many is essential is its singularity and other urban design qualities – Imageability location, in relationship to its context, legibility, enclosure, human scale, Imageability is the quality of a place background, and the city at large. transparency, linkage, complexity, that makes it distinct, recognizable, Landmarks are a principle of urban and coherence – and is in some way and memorable. A place has high design because they act as visual the net effect of these qualities. imageability when specific physical termination points, orientation points, Places that rate high on these elements and their arrangement and points of contrast in the urban qualities are likely to rate high on capture attention, evoke feelings, and setting. Tunnard and Pushkarev imageability as well – the create a lasting impression. (1963, p. 140) attribute even greater neighborhoods of Paris or San importance to landmarks, saying “A Francisco, for example. However, Discussion landmark lifts a considerable area places that rate low on these qualities Kevin Lynch (1960) defines around itself out of anonymity, giving may also evoke strong images, imageability as a quality in a physical it identity and visual structure.” though ones that people may prefer object that gives it a high probability Imageability is related to “sense of to forget. Although the strength of of evoking a strong image in any place.” Gorden Cullen (1961, p. 152) the image a place evokes, whether given observer: “It is that shape, elaborates on the concept of a “sense positive or negative, is itself of color, or arrangement which of place,” asserting that a interest, urban designers focus on the facilitates the making of vividly characteristic visual theme will strength of positive images in identified, powerfully structured, contribute to a cohesive sense of discussing imageability and sense of highly useful mental images of the place, and will inspire people to enter place. environment.” A highly imageable and rest in the space. Jan Gehl Legibility city is well formed, contains distinct (1987, p. 183) explains this parts, and is instantly recognizable to phenomena using the example of Legibility refers to the ease with anyone who has visited or lived famous Italian city squares, where which the spatial structure of a place there. It plays to the innate human “life in the space, the climate, and can be understood and navigated as ability to see and remember patterns. the architectural quality support and a whole. The legibility of a place is It is one whose elements are easily complement each other to create an improved by a street or pedestrian identifiable and grouped into an unforgettable total impression.” network that provides travelers with overall pattern. When all factors manage to work a sense of orientation and relative Landmarks are a key component of together to such pleasing ends, a location and by physical elements imageability. The term “landmark” feeling of physical and psychological that serve as reference points.

40 Discussion ambiguous. A regular grid of streets embraced the concept of visual makes it easy for people to navigate termination. Andres Duany and The dictionary defines legibility as even when they are unfamiliar with a Elizabeth Plater-Zyberk (1992), “possible to read or decipher, plainly place, although it does not provide a pioneers of this movement, say that discernible, or apparent.” As way of distinguishing one block from visual terminations focus the described by Kevin Lynch in his another. An irregular pattern of community, as well as put a definite classic, The Image of the City, streets, in which blocks are of end point to streets to keep them legibility is the apparent clarity of the irregular length and compass from going on forever. On a large cityscape, the “ease by which its orientation changes from block to scale, visual termination points can parts can be recognized and can be block, may increase the difficulty of include large civic buildings, organized into a coherent pattern.” navigating and learning the network, prominent landmarks, or elements of Lynch suggests that when faced with although it distinguishes each block nature. On a smaller, neighborhood a new place, people automatically with different lengths and scale, visual termination can be create a mental map that divides the orientations. The street network thus created by the use of gazebos, bends city into paths, edges, districts, works together with other elements in the roads, or other small-scale nodes, and landmarks. Places with of the physical environment to elements. Allan Jacobs (1993, p. strong edges, distinct landmarks, and determine the legibility of a place. 297) says of streets, “Since they busy nodes allow people to form Signage, in particular, helps to have to start and stop somewhere, detailed and relatively accurate distinguish one point from another these points should be well marked.” mental maps. Conversely, a city that and to orient and direct a traveler He argues that clearly marked end has no definite edges, nodes, or through the network. Landmarks, points both serve as reference points visually interesting features, will be which have an important influence on and give a sense of definition to an difficult to make sense of and to imageability, also play an important area. remember. Legibility facilitates role in mental maps and thus help to To our knowledge, only visual wayfinding, the process by which increase the legibility of a place. assessment study has attempted to people move successfully through the Visual termination and deflection measure legibility, this in connection physical environment to reach a points also contribute to legibility. with natural rather than urban desired destination, determining a Visual termination creates a “focal landscapes (Herzog and Leverich route between two points, choosing point, the vertical symbol of 2003). Legibility was highly an alternate route when the primary congregation” (Cullen 1961, p. 26). correlated with another perceptual route is impassable, navigating along “…in the fertile streets.. it is the focal quality, coherence. The hypothesized a route, and learning a new spatial point which crystallizes the situation, relationship to landmarks proved to environment. which confirms ‘this is the spot’, ‘Stop be weak. The layout of the street network has looking, it is here.’” Recently, the an important influence on legibility, New Urbanism movement has although the influence is sometimes

41 Enclosure memorable—an invitation to enter a At low suburban densities, building place special enough to warrant masses become less important in Enclosure refers to the degree to boundaries. Jacobs and Appleyard defining space, and street trees which streets and other public spaces (1987) speak of the need for assume the dominant role. Rows of are visually defined by buildings, buildings to “define or even enclose trees on both sides of a street can walls, trees, and other vertical space—rather than sit in space.” humanize the height-to-width ratio. elements. Spaces where the height Richard Hedman (1984) refers to Henry Arnold (1993) explains that of vertical elements is proportionally certain arrangements of buildings trees define space both horizontally related to the width of the space creating intensely three dimensional and vertically. Horizontally they do between them have a room-like spaces. so by visually enclosing or completing quality. In an urban setting, enclosure is an area of open space. Vertically Discussion formed by lining the street or plaza they define space by creating an airy with unbroken building fronts of ceiling of branches and leaves. Outdoor spaces are defined and roughly equal height. The buildings Unlike the solid enclosure of shaped by vertical elements, which become the "walls" of the outdoor buildings, tree lines depend on visual interrupt viewers' lines of sight. A room, the street and sidewalks suggestion and illusion. Street space sense of enclosure results when lines become the "floor," and if the will seem enclosed only if trees are of sight are so decisively blocked as buildings are roughly equal height, closely spaced. Properly scaled, walls to make outdoor spaces seem room- the sky projects as an invisible and fences can also provide spatial like. Gordon Cullen (1961) states ceiling. Buildings lined up that way definition in urban and suburban that “Enclosure, or the outdoor room, are often referred to as "street walls." settings. Kevin Lynch recommended is, perhaps, the most powerful, the Alexander et al. (1977, pp. 489-491) walls and fences that are either low most obvious, of all the devices to state that the total width of the or over six feet tall. instill a sense of position, of identity street, building-to-building, should Visual termination points may also with the surroundings….it embodies not exceed the building heights in contribute to a sense of enclosure. the idea of hereness…” Christopher order to maintain a comfortable Andres Duany and other New Alexander et al. (1977, p. 106) say feeling of enclosure. Allan Jacobs Urbanists advocate closing vistas at that “An outdoor space is positive (1993) is more lenient in this regard, street ends with prominent buildings, when it has a distinct and definite suggesting that the proportion of monuments, fountains, or other shape, as definite as the shape of a building heights to street width architectural elements as a way of room, and when its shape is as should be at least 1:2. Other achieving enclosure in all directions important as the shapes of the designers have recommended (Duany and Plater-Zyberk 1992). buildings which surround it. Likewise, proportions as high as 3:2 and as low When a street is not strongly defined Allan Jacobs (1993) says that people as 1:6 for a sense of enclosure. by buildings, focal points at its ends react favorably to fixed boundaries as can maintain the visual linearity of something safe, defined, and even

42 the arrangement. Similarly, the Human Scale floors should spread out and upper layout of the street network can floors step back before they ascend, Human scale refers to a size, texture, influence the sense of enclosure. A giving human-scale definition to and articulation of physical elements rectilinear grid with continuous streets and plazas. Richard Hedman that match the size and proportions streets creates long sight lines that (1984) emphasizes the importance of of humans and, equally important, may offset the sense of enclosure articulated architecture and belt correspond to the speed at which created by the buildings and trees courses and cornices on large humans walk. Building details, that line the street. Irregular grids buildings to help define street space pavement texture, street trees, and may create visual termination points and scale. Several authors suggest street furniture are all physical that help to enclose a space; cul-de- that the width of buildings, not just elements contributing to human sacs, for example, tend to create the height, defines human scale. For scale. more sense of enclosure than human scale, building widths should through streets. Discussion not be out of proportion with building Enclosure is eroded by breaks in the heights, as are so many buildings in The urban design glossary for the continuity of the street wall, that is, the suburbs. In what was billed as City of Seattle defines human scale breaks in the vertical elements such the first of its kind, Stamps (1998) as “the quality of a building that as buildings or tree rows that line the used a visual assessment survey to includes structural or architectural street. Breaks in continuity that are explore perceptions of architectural components of size and proportions occupied by non-active uses create mass. The most important that relate to the human form and/or dead spaces that further erode determinant was the cross sectional that exhibits through its structural or enclosure; vacant lots, parking lots, area of buildings, second was the architectural components the human driveways, and other uses that do not amount of fenestration, and third was functions contained within.” Modest generate human activity and the amount of façade articulation and sized buildings, narrow streets and presence are all considered dead partitioning. small spaces create an intimate spaces. Large building setbacks are Human scale can also be defined by environment, and the opposite for another source of dead space. human speed. Jane Holtz Kay (1997) large buildings, streets and spaces. Alexander et al. (1997) say “building argues that today, far too many Urban designers offer differing setbacks from the street, originally things are built to accommodate the definitions of human scale. Alexander invented to protect the public welfare bulk and rapid speed of the et al. (1977) state that any buildings by giving every building light and air, automobile; we are “designing for 60 over four stories tall are out of have actually helped greatly to mph.” When approached by foot, human scale. Lennard and Lennard destroy the street as social space.” these things overwhelm the senses, (1987) set the limit at six stories. creating disorientation. For example, Hans Blumenfeld (1953) sets it at large signs with large lettering are three stories. In taller buildings, designed to be read by high-speed Roger Trancik (1986) says that lower

43 motorists. For pedestrians, small such as clock towers to moderate the edge need only be imagined, not signs with small lettering are much scale of buildings and streets. actually seen. Allan Jacobs (1993) more comfortable. says that streets with many Transparency Personal interaction distances play a entryways contribute to the role in designing for the human scale. Transparency refers to the degree to perception of human activity beyond Jan Gehl (1987) designates these which people can see or perceive the street, while those with blank distances as: what lies beyond the edge of a street walls and garages suggest that or other public space and, more people are far away. Even blank Intimate distance 0-1.5 ft specifically, the degree to which walls may exhibit some transparency Personal distance 1.5-4.5 ft people can see or perceive human if overhung by trees or bushes, Social distance 4.5-12 ft activity beyond the edge of a street providing signs of habitation. Henry Public distance >12 ft or other public space. Physical Arnold (1993) tells us that trees with elements that influence transparency high canopies create "partially According to Alexander et al. (1977), include walls, windows, doors, transparent tents," affording a person's face is just recognizable at fences, landscaping, and openings awareness of the space beyond while 70 feet, a loud voice can just be into midblock spaces. still conferring a sense of enclosure. heard at 70 feet, and a person's face By contrast, small trees in most Discussion is recognizable in portrait-like detail urban settings work against up to about 48 feet. These set the Taken literally, transparency is a transparency (Arnold 1993). limits of human scale for social material condition that is pervious to Transparency is most critical at the interaction. light and/or air, an inherent quality of street level, because this is where the Street trees can moderate the scale substance as in a glass wall. As an greatest interaction occurs between of tall buildings and wide streets. element of design, transparency has indoors and outdoors. William H. According to Henry Arnold (1993), two meanings: an ability to see Whyte (1988) opined that if a blank where tall buildings or wide streets beyond the street edge and signs of wall index were ever computed, as would intimidate pedestrians, a habitation beyond the street edge. A the percentage of blockfront up to canopy of leaves and branches allows classic example of transparency is a 35-foot height that is blank, it would for a simultaneous experience of the shopping street with display windows show that blank walls have become smaller space within the larger that invite passersby to look in and the dominant feature of cityscapes. volume. He posits that where streets then come in to shop. Blank walls Linkage are over 40 feet wide, additional rows and reflective glass buildings are of trees are needed to achieve human classic examples of the elements that Linkage refers to physical and visual scale. Hedman (1984) recommends reduce transparency. connections from building to street, the use of other small-scale elements But transparency can be subtler than building to building, space to space, this. What lies behind the street or one side of the street to the other

44 which tend to unify disparate breaking up the gridiron with barriers Discussion elements. Tree lines, building and diverters. Amos Rapoport (1990) explains the projections, marked crossings all Linkages between the street and fundamental properties of create linkage. Linkage can occur surrounding buildings are also complexity. Complexity is related to longitudinally along a street or important and may be psychological the number noticeable differences to laterally across a street. as well as physical. Maintenance of which a viewer is exposed per unit sight lines and sidewalk connections Discussion time. Human beings are most are obvious ways to provide this kind comfortable receiving information at Linkages can be defined as features of linkage, but it can also be provided perceivable rates. Too little that promote the interconnectedness in more subtle ways. For example, information produces sensory of different places and that provide Henry Arnold (1993) advocates the deprivation, too much creates convenient access between them. use of trees for linkage: continuous sensory overload. Rapoport contrasts Linkage is closely associated with the tree rows can psychologically connect the complexity requirements of concept of connectivity, as both are places at either end, and tree pedestrians and motorists. Slow concerned with the ease of patterns that reflect or amplify moving pedestrians require a high movement in an area and depend on building geometry can psychologically level of complexity to hold their the relationships between paths and link buildings to the street. As Roger interest. Fast moving motorists will nodes. Jacobs (1993) recommends Trancik (1986) puts it: “Urban design find the same environment chaotic. urban street connections at most is concerned with the question of The commercial strip is too complex every 300 feet. Alexander et al. making comprehensible links between and chaotic at driving speeds, yet (1977) give similar advice, discrete things.” In this way, the due to scale, yields few noticeable suggesting pedestrian road crossings concept of linkage is closely related differences at pedestrian speeds. every 200 or 300 feet. They to the concept of legibility. The environment can provide low advocate the use of a separate Complexity levels of usable information in three pedestrian-only network running ways: elements may be too few or orthogonal to the street grid to Complexity refers to the visual too similar; elements, though maximize pedestrian accessibility. richness of a place. The complexity numerous and varied, may be too Duany and Plater-Zyberk (1992) of a place depends on the variety of predictable for surprise or novelty; or generally limit the size of blocks to the physical environment, specifically elements, though numerous and 230 by 600 feet to ensure reasonable the numbers and kinds of buildings, varied, may be too unordered for travel distances. On the other hand, architectural diversity and comprehension. Pedestrians are apt Appleyard (1981) argues against too ornamentation, landscape elements, to prefer streets high in complexity, much connectivity in residential areas street furniture, signage, and human since they provide interesting things since through traffic can erode a activity. to look at: building details, signs, sense of community, and suggests people, surfaces, changing light

45 patterns and movement, signs of Complexity is one perceptual quality more inviting, and help create a habitation. As Jan Gehl notes in his that has been measured extensively sense of place. Gordon Cullen (1961, book Life Between Buildings, an in visual assessment studies. It has p. 151) calls advertisement signs “the interesting walking network will have been related to changes in texture, most characteristic, and, potentially, the “psychological effect of making width, height, and setback of the most valuable, contribution of the the walking distance seem shorter,” buildings (Elshestaway 1997). It has twentieth century to urban scenery.” by virtue that the trip is “divided been related to building shapes, When these signs are lit up at night, naturally, into manageable stages.” articulation, and ornamentation the result can be spectacular. Complexity results from varying (Stamps 1999; Heath et al. 2000). However, signage must not be building shapes, sizes, materials, Other elements of the built allowed to become chaotic and colors, architecture, and environment also contribute to unfriendly to pedestrian traffic. ornamentation. According to Jacobs complexity. According to Henry Nasar (1987) reports that people and Appleyard (1987), narrow Arnold (1993), one function of trees prefer signage with moderate rather buildings in varying arrangements is to restore the rich textural detail than high complexity—measured by add to complexity, while wide missing from modern architecture. the amount of variation among signs buildings subtract. Allan Jacobs Light filtered through trees gives life in location, shape, color, direction, (1993) refers to the need for many to space. Manipulation of light and and lettering style. Allan Jacobs different surfaces over which light is shade transforms stone, asphalt, and (1993) uses Hong Kong signage as constantly moving in order to keep concrete into tapestries of sunlight an example of complexity to the point eyes engaged. Tony Nelessen (1994) and shadow. Allan Jacobs (1993) of chaos. asserts that “Variations on basic similarly values to the constant The presence and activity of people patterns must be encouraged in order movement of branches and leaves, add greatly to the complexity of a to prevent a dull sameness. If a and ever-changing light that play on, scene. This is true not only because particular building or up to three through, and around them. Street people appear as discrete “objects,” buildings are merely repeated, the furniture also contributes to the but because they are in constant result will be boring and mass complexity of street scenes. Jacobs motion. Jan Gehl (1987, p. 25) produced.” Variation can be (1993) states that pedestrian-scaled explains that “people are attracted to incorporated into the building streetlights, fountains, carefully other people. They gather with and orientation plan or building set-back thought out benches, special paving, move about with others and seek to line, allowing for varied building even public art, combine to make place themselves near others. New frontage instead of monotonous, regal, special places. activities begin in the vicinity of straight building frontage. Numerous Signage is a major source of events that are already in progress.” doors and windows produce complexity in urban and suburban Allan Jacobs, in the course of his complexity as well as transparency. areas. If well done, signs can add world wide travels, found that the visual interest, make public spaces most popular streets were ones that

46 contained “sidewalks fairly cluttered Discussion Richard Hedman (1984, p. 29) warns with humans and life,” calling them that when every building seeks to be Coherence refers to the orderly an “attractive obstacle courses” that a unique statement and the center of arrangement of physical elements in never failed to entertain. attention, there is an unexpected the environment (Kaplan and Kaplan Complexity can also arise at a larger effect—“instead of providing an 1989, p. 54). Allan Jacobs (1993, p. scale from the pattern of exciting counterpoint, the addition of 287) describes coherence in development. According to each new and different building architecture as follows: “...buildings Christopher Alexander (1965), intensifies the impression of a on the best streets will get along with organically developed older cities nervous, irritating confusion.” He each other. They are not the same, have complex "semi-lattice" goes on to list multiple features of but they express respect for one structures, while new planned buildings that, when repeated, can another, most particularly in respect developments have simple "tree-like create visual unity: building to height and the way they look.” structures." Integration of land uses, silhouettes, spacing between According to Henry Arnold (1993), housing types, activities, buildings, setbacks from street, complexity of architecture of earlier transportation modes, and people proportions of eras was given coherence by creates diversity, and that in turn windows/bays/doorways, massing of common materials, handcrafted adds to complexity (Gehl 1987). building form, location of entryways, details, and reflections of human use. Jane Jacobs (1961, p. 161) describes surface material and finish, shadow Because these are absent from diversity as a mixture of commercial, patterns, building scale, style of today's architecture, landscaping residential, and civic uses in close architecture, and landscaping. becomes critical for creating a sense proximity to each other, creating While often presented as opposites, of visual unity; shade trees planted human traffic throughout day and coherence and complexity represent close together result in an night, and subsequently benefiting distinct perceptual dimensions. uninterrupted pattern of light and the safety, economic functioning, and Visual preference surveys show that shade, unifying a scene. At the city appeal of a place. viewers do not appreciate massive level, coherence takes the form of doses to unstructured information. Coherence orderly density patterns and People like complexity, but not the hierarchies of communal spaces Coherence refers to a sense of visual unstructured complexity of the (Alexander et al 1977). Nikos order. The degree of coherence is commercial strip. Scenes with high Salingaros, applying mathematical influenced by consistency and complexity and low coherence tend to principles to the urban setting, complementarity in the scale, be least liked, causing Herzog et al. concludes, “Geometrical coherence is character, and arrangement of (1982) to conclude that, "High an identifiable quality that ties the buildings, landscaping, street complexity urban areas must also be city together through form, and is an furniture, paving materials, and other highly coherent.” In one such survey, essential prerequisite for the vitality physical elements. Nasar (1987) found that people of the urban fabric.”

47 preferred signage that is moderate complex and highly coherent. Generalizing across many surveys, Kaplan and Kaplan (1989, p. 54) deem scenes of low complexity and high coherence as "boring," scenes of high complexity and low coherence as "messy," but scenes of high complexity and high coherence as "rich and organized.” It is important to note that coherence does not imply mindless repetition or blandness, rather continuity of design and thematic ordering.

48 Appendix 3: Operational Definitions of Physical Features

Variable Long Name Variable Type Counting Criteria Measurement Protocol

pass camera or Count individual courtyards, plazas, parks that the camera passes on either side of the street within 50 feet, or that are within 50 feet from the camera. Large parks that occupy a whole block will still courtyards/plazas/parks - both sides count both sides count as one park.

Indicate the presence of an arcade. An arcade will be defined as a covered passageway which allows the passageway to be protected from rain and direct sun, while retaining the dummy advantages of an outdoor space. Count arcades regardless if the camera is inside or outside arcade – same side 1 = yes 0 = no the arcade.

Count number of landmarks. Landmark must be at least 20% of the screen height. Landmarks are defined as a building or structure that stands out from the background buildings. The 20% screen structure should be prominent and /or well known that it could be plausibly used as a landmarks - both sides count height, both sides reference point for orientation and for providing directions to visitors.

types of landmarks text List landmarks counted.

Count views of mountain ranges, bodies of water, and other man made features that incorporate the surrounding environment (a marina) that would serve as natural landmarks. The major landscape feature should be prominent and/or well known that it could be plausibly 20% screen used as a reference point for orientation and for providing directions to visitors. The feature major landscape features - both sides count height, both sides should occupy at least 20% of the screen height.

types of major landscape features text List major landscape features counted.

49 Variable Long Name Variable Type Counting Criteria Measurement Protocol

Indicate the presence of memorable architecture. This is defined as more than just well designed buildings. Memorable architecture implies that the scene as a whole contains architecture that makes the scene prominent and/or well known. The scene as a whole could dummy be plausibly used as a reference point for orientation and for providing directions to visitors. A memorable architecture 1 = yes 0 = no counted buildings well known landmark can serve as memorable architecture

buildings with memorable architecture text List buildings that contribute to the scene having memorable architecture.

20% screen height, both sides; Indicate the presence of distinctive signage. A sign is distinctive if it is prominent and/or well or passed, or known that it could be plausibly used as a reference point for orientation and for providing dummy within 50 feet, directions to visitors. The occurrences of distinctive signage should occupy at least 20% of distinctive signage 1 = yes 0 = no same side the screen height or be within 50 feet of the camera.

occurrences of distinctive signage text List instances where distinctive signage occurs.

Indicate the number of directions in which the camera can see far into the distance. Maximum number is 3 (right, left, front). "Far into the distance" will be defined as seeing at least 1000 long sight lines count 1000 feet ahead feet into the distance.

Indicate whether the street the camera travels along is terminated with a building or feature dummy that blocks distant views. The feature that blocks distant views must occupy at least 20% of terminated vista 1 = yes 0 = no 20% screen height screen height.

Estimate how far the camera has traveled in relation to the end of the block. Instances where a street that intersects with the opposite side of the street but not with the side where the camera is traveling (a t-intersection) will not be considered as the end of the block if the intersecting street appears to be minor (less than 2 marked lanes, no signalization, no marked progress toward next intersection proportion min .05 crosswalks). Enter 0.05 if next intersection is not visible. Use 0.10 intervals otherwise.

50 Variable Long Name Variable Type Counting Criteria Measurement Protocol

Estimate how far the camera has traveled in relation to the most distant feature seen. Enter proportion of distance walked versus distance 0.05 if distance beyond seems infinite. Use 0.10 intervals. Use the same value as progress visible proportion min .05 toward next intersection if farthest distance visible is the end of the block.

Count visible street connections. The camera must be able to see down the street or street connections to elsewhere count pedestrian way to count as a connection to elsewhere.

20% screen Count buildings along street and in the distance that occupy at least 20% of screen height. number of buildings – both sides count height, both sides Large structures that are subdivided count as one building.

Count different land uses observed on both sides of street. Land use distinctions are civic/community, residential, lodgings, office, medical, retail (includes restaraunts and shops), entertainment, transit station, and park. Parking, even in a structure, will not be considered a 20% screen land use. Only count land uses from features that occupy at least 20% of screen height and/or number of land uses - both sides count height, both sides from buildings that have been counted.

types of land uses text List land uses counted.

counted buildings, and fronting along Estimate the proportion of the street fronted by buildings that is fronted by buildings that are street and passed historic. Architecture that can be determined to have originated from the World War II era or proportion of historic building frontage - both or 500 feet ahead, before will be considered as historic. Relevant frontage is defined as the total distance the sides proportion both sides camera travels plus an additional 500 feet ahead from the end of the video clip.

types of historic buildings text Identify the buildings in the scene that are historic.

51 Variable Long Name Variable Type Counting Criteria Measurement Protocol

counted buildings, and fronting along Count buildings whose use can be determined by building features. For example, a church street and passed can be identified by a steeple. Stores can be identified by signs that can be easily read in the number of buildings with identifiers - both or 500 feet ahead, video clip. If a building has been subdivided by several occupants, only count the building as sides count both sides indentifiable if a majority of the occupants' uses can be determined by building features.

counted buildings, and fronting along street and passed Determine the building frontage whose uses can be determined by building features. Relevant or 500 feet ahead, frontage is defined as the total distance the camera travels plus an additional 500 feet ahead proportion of building frontage with identifiers proportion both sides from the end of the video clip.

at least 1 counted dummy building from Indicate whether buildings appear to have been built at different time periods. At least one various building ages 1 = yes 0 = no different period counted building must appear to be built in a different time period.

Count different primary building materials for buildings that have been counted. Glass counts number of primary building materials count counted buildings as a building material only if it makes up the entire building.

types of primary building materials text List counted building materials.

Count different dominant building colors for buildings that have been counted. If the roof color number of dominant building colors count counted buildings of a building is different than the building, the roof color will count as an accent color.

dominant building colors text List counted dominant building colors for buildings that have been counted.

52 Variable Long Name Variable Type Counting Criteria Measurement Protocol counted buildings, objects that occupy 20% of Count number of accent colors. Accent colors contrast the dominant building colors and can screen height or come from street furniture, awnings, business signs, and building trim. Accent colors will be within 50 feet, counted only from objects that meet one of the counting conventions. The object must at least number of accent colors - both sides count both sides occupy 20% of the screen height or be within 50 feet from the camera.

accent colors text List counted accent colors.

Count building projections (such as porches, stoops,marquees, decks, balconies, window at least 5 feet, bays, etc…) that project at least 5 feet from the building and are from buildings that have been passed or 50 feet counted that front the street, and that are passed or within 50 from the camera at the end of building projections – same side count ahead, same side the clip.

Count sets of doors that the camera passes or that are within 50 feet from the camera on the passed or 50 feet same side of the street. Do not assume the location of doors or count doors seen in window visible sets of doors – same side count ahead, same side reflections.

Count number of recessed doorways of counted visible doorways. Doorways are recessed if visible recessed doors – same side count counted doors they are setback at least 3 feet from the building façade.

proportion of counted sets of door that are recessed proportion Divide visible recessed doors by visible sets of doors. fronting along street, and passed or 50 feet ahead, and set back no Estimate the proportion of the first floor of buildings that front the street on the same side that proportion of first floor façade that has more than 50 feet are passed or within 50 feet from the camera at the end of the clip that is window. Use 0.10 windows - same side proportion same side intervals.

53 Variable Long Name Variable Type Counting Criteria Measurement Protocol fronting along street, and passed or 50 feet ahead, and set back no Estimate the proportion of the entire surface of buildings that front the street on the same side proportion of entire façade that has windows - more than 50 feet that are passed or within 50 feet from the camera at the end of the clip that is window. Use same side proportion same side 0.10 intervals. at least 80% of counted buildings fronting along Indicate whether windows have common proportions. Common window proportions occur street and passed when at least 80% of windows on all buildings are predominantly vertical or horizontal and dummy or 500 feet ahead, have similar architectural trim. If a building on one of the sides has no window than there is no common window proportions - both sides 1 = yes 0 = no both sides common window proportions.

counted buildings fronting along street and passed or 50 feet ahead, Count number of awnings or overhangs on buildings that have been counted on both sides of awnings or overhangs – both sides count both sides street and that are passed or are within 50 feet from the camera at the end of the clip.

counted buildings fronting along Estimate proportion of building frontage that have been counted on the same side that front street and passed the street and are within 50 feet from the camera at the end of the video clip with belt courses or 50 feet ahead, or other visual interruptions to building height. One storey buildings should be considered as height interruptions – same side proportion same side height interrupted. Use 0.10 intervals.

Count buildings that have been counted whose shape is not a simple rectangular box. Pitched roofs on buildings that are viewed at an angle and make the building look non-rectangluar do number of buildings with non-rectangular count as non-rectangular. Building roof trim that makes variations in an otherwise simple silhouettes count counted buildings rectangular shape do also count as non-rectangular.

proportion of buildings with non-rectangular silhouettes proportion Divide number of buildings with non-rectangular silhouettes by number of counted buildings. at least 80% of counted buildings fronting along Indicate the presence of common architectural styles. Common architectural styles occurs street and passed when at least 80% of the counted buildings that front the street and that have been passed or dummy or 500 feet ahead, are within 500 feet from the camera at the end of the video clip use similar architectural styles common architectural style - both sides 1 = yes 0 = no both sides and/or have consistent building trim and roof pitch.

54 Variable Long Name Variable Type Counting Criteria Measurement Protocol at least 80% of counted buildings fronting along Indicate the presence of common building materials. Common building materials occurs when street and passed at least 80% of the counted buildings that front the street and that have been passed or are dummy or 500 feet ahead, within 500 feet from the camera at the end of the video clip use the same primary building common materials - both sides 1 = yes 0 = no both sides material. Determine proportion of street frontage that has active uses . Active uses are defined as shops, restaraunts, public park, and other uses that generate significant pedestrian traffic. counted buildings Inactive uses include blank walls, parking lots, vacant lots, abandoned buildings, and office fronting along with no apparent activity. In Regards to residential uses, when the density appears to be more street and passed than 10 units per acre, assume the land use to be active. The street frontage will be defined or 50 feet ahead, as the total distance traveled by the camera plus an additional 50 feet ahead from the end of proportion active uses – same side proportion same side the video clip.

counted buildings Determine proportion of street that is occupied by a continuous wall or facade adjacent to the fronting along sidewalk. Facades set back by parking or lawn and driveways do not count as street wall. street and passed Intersecting streets and ends of blocks however should not count against street wall. The or 500 feet ahead, street will be defined as the total distance the camera travels plus an additional 500 feet proportion street wall – same side proportion same side ahead from the end of the video clip.

counted buildings Determine proportion of street that is occupied by a continuous wall or facade adjacent to the fronting along sidewalk. Facades set back by parking or lawn do not count as street wall. Driveways also do street and passed not count as street wall. Intersecting streets and ends of blocks however should not count or 500 feet ahead, against street wall. The street will be defined as the total distance the camera travels plus an proportion street wall – opposite side proportion opposite side additional 500 feet ahead from the end of the video clip. Indicate the number of sides of the street that are enclosed. Maximum number is 3 (front, at least 80% of same side, opposite side). A side is considered enclosed if 80% of the frontage on that side is frontage passed blocked by buildings or other featrues that are opaque at street level whether or not they front or 500 feet ahead along the sidewalk. If the street has a terminated by a vista then the front is enclosed. that is blocked, Relevant frontage is defined as the total distance the camera travels plus an additional 500 enclosed sides count both sides feet ahead from the end of the video clip.

passed on the average building setback from sidewalk or same side or 50 ft Estimate the average setback of buildings from the sidewalk or travel path on the same side. travel path - same side dimension ahead Buildings that front directly on the sidewalk have a setback of 0. no more than 30% variance for Indicate whether buildings that have been counted, that front along the street, and are within buildings fronting 500 feet from the camera at the end of the video clip have a common setback. Common along street and setbacks occure when all building setbacks vary no more than 30%. Recessed courtyards and dummy passed or 500 feet other small breaks in street wall that can be determined as part of a building do not negate the common setbacks – both sides 1 = yes 0 = no ahead, both sides presense of common setbacks.

55 Variable Long Name Variable Type Counting Criteria Measurement Protocol Estimate average building height of buildings on same side of street based on proportion of street fronted by each building. Only estimate building heights for buildings that have been counted, that front along the street, and are within 500 feet from the camera at the end of the video clip. Use 0 if there are no buildings that front along street. Only estimate the height of passed or 500 feet building that you can see if the camera does not pan the entire height of the building. Assume building height – same side dimension ahead, same side that the height for 1 typical storey is 10 feet.

Estimate average width of buildings on same side. Only estimate the ratio for buildings that have been counted, that front along the street, and are within 500 feet from the camera at the passed or 500 feet end of the video clip. Divide average height of buildings on same side by average width of building height to width ratio – same side ratio ahead, same side buildings on same side.

Estimate average building height of buildings on opposite side of street. Only estimate building heights for buildings that have been counted, that front along the street, and are within 500 passed or 500 feet feet from the camera at the end of the video clip. Use 0 if there are no buildings that front ahead, opposite along street. Only estimate the height of building that you can see if the camera does not pan building height – opposite side dimension side the entire height of the building. Assume that the height for 1 typical storey is 10 feet. no more than 30% variance for buildings fronting along street and Indicate whether buildings have common buildiing heights. Common building heights occur dummy passed or 500 feet when the height of all buildings that have been counted, that front along the street, and are common building heights – both sides 1 = yes 0 = no ahead, both sides within 500 feet from the camera at the end of video clip varies no more than 30%. no more than 30% variance for buildings fronting along street and Indicate whether buildings have common building masses. Common mass occurs when when dummy passed or 500 feet the mass of all buildings that have been counted, that front along the street, and are within common building masses – both sides 1 = yes 0 = no ahead, both sides 500 feet from the camera at the end of video clip varies no more than 30%.

Estimate street width. Street width includes frontage roads and/or parking aisles. Assume that a typical parking lane is 8 feet and a typical travel lane is 12 feet. If the median between a one average of passed way pair is more than 50 feet wide, treat each couplet as seperate streets. Only consider the street width dimension or 50 feet ahead width of the adjacent half street.

average of passed median width dimension or 50 feet ahead Estimate median width if one is present. Medians should be raised in order to be considered.

56 Variable Long Name Variable Type Counting Criteria Measurement Protocol

average of passed or 50 feet ahead, sidewalk width – same side dimension same side Estimate total sidewalk width. Estimate average if sidewalk width varies.

Street width is defined as total width from building face to building face. If building face to building face distance varies, estimate an average. Compute average of building heights of both sides of street. Divide by total width of street including street, median, sidewalks, and building height to street width ratio ratio average setback from sidewalk.

average of passed or 50 feet ahead, sidewalk clear width - same side dimension same side Estimate width of sidewalk with no obstructions.

average of passed or 50 feet ahead, Estimate width from outside clear width to moving cars. (distance between moving cars and buffer width - same side dimension same side the portion of the sidewalk where pedestrians are most likely to walk).

Count number of different paving materials for the street, the sidewalk on same side, and passed or 50 feet surfaces connected to the sidewalk on same side. Paving material categories are asphalt, number of paving materials count ahead concrete, colored concrete, brick, paver, and aggregate.

types of paving materials text List counted paving materials.

Indicate the presence of a textured sidewalk. Textured sidewalks or streets are composed of materials that have patterns (brick, pavers, stamped asphalt, patterned or stameped concrete). The patterned materials usually resemble brick and are used to visually break up dummy passed or 50 feet sidewalk or street. A sidewalk or street will be considered textured if at least 50% of the textured sidewalk surface – same side 1 = yes 0 = no ahead, same side surface is textured.

57 Variable Long Name Variable Type Counting Criteria Measurement Protocol

Indicate the presence of a textured street. Textured sidewalks or streets are composed of materials that have patterns (brick, pavers, stamped asphalt, patterned or stameped concrete). The patterned materials usually resemble brick and are used to visually break up dummy passed or 50 feet sidewalk or street. A sidewalk or street will be considered textured if at least 50% of the textured street surface 1 = yes 0 = no ahead surface is textured.

Rate pavement condition on a 1-5 Likert scale taking note of visible cracks,discoloration, passed or 50 feet patches, presence of weeds, etc... Rate the condition of the sidewalk on the same side and pavement condition rating ahead the street.

pavement condition explanation text Explain pavement condition rating.

passed or 50 feet Rate debris condition of pavement on a 1-5 Likert scale taking note of dirt, leaves, and trash. debris condition rating ahead Rate the condition of the sidewalk on the same side and the street.

debris condition explanation text Explain debris condition rating.

passed or 50 feet parked cars – same side count ahead, same side Count parked cars on same side that are within 50 feet from the camera.

Estimate proportion of street frontage on same side with parked cars. Make deductions for proportion of street with parked cars – same passed or 50 feet occupied parking spaces that are extra long. The relevant frontage is defined as the total side proportion ahead, same side distance the camera travels plus an additional 50 feet ahead from the end of the video clip.

58 Variable Long Name Variable Type Counting Criteria Measurement Protocol

passed or 50 feet Count number of cars that pass camera on either side of street or that are within 50 feet from moving cars – both sides count ahead, both sides the camera.

speed measurement Estimate speed of moving cars using 5 mph intervals.

traffic to street width ratio ratio Divide the number of moving cars on both sides by the street width.

passed or 50 feet moving cyclists - both sides count ahead, both sides Couunt moving bicyclists on either side of street or that are within 50 feet from the camera.

Count curb extensions that are passed or that are within 50 feet from the camera. Curb extensions are extensions of the sidewalk into the street to facilitate shorter distances for passed or 50 feet pedestrians to cross and to slow down oncoming vehicular traffic. They can occur midblock or curb extensions – same side count ahead, same side at intersections.

passed or 50 feet Count midblock crossings that are passed or that are within 50 feet from the camera. Midblock midblock crossings count ahead, same side crossings are marked crossings for pedestrians that do not occur at street intersections.

passed or 50 feet Count open passageways into street wall (such as alleys) that are passed or are within 50 feet midblock passageways – same side count ahead, same side from the camera.

59 Variable Long Name Variable Type Counting Criteria Measurement Protocol

dummy overhead utilities 1 = yes 0 = no Indicate the presence of overhead utility lines.

Count each type of tree, bush, and visible groundcover. Distinguish between trees in natural 20% screen settings vs. trees in wells or landscaped beds, evergreen trees vs deciduous, small trees vs. height, both sides; tall trees. Note the occurences of bushes and/or hedges and turf. Only count landscape or within 50 feet, elements that either occupy at least 20% of screen height or are within 50 feet from the number of landscape elements - both sides count same side camera.

type of landscape elements text List counted landscape elements.

dummy passed or 50 feet Indicate the presence of a landscaped median. The median should be landscaped for more landscaped median 1 = yes 0 = no ahead than 50% for the median that is passed and ahead 50 feet from the end of the clip.

20% screen number of trees – both sides count height, both sides Count total number of trees on both sides that occupy at least 20% of screen height.

at least 10 square feet and within 50 Count number of trees that are in tree wells, well landscaped adorned planting beds, or well trees in wells or landscaped beds - same side count feet, same side landscaped fenced areas that are at least 10 square feet and within 50 feet from the camera.

Estimate proportion of sidewalk that is shaded by trees. The relevant sidewalk is defined as proportion of sidewalk shaded by trees – passed or 50 feet the total distance the camera travels plus an additional 50 feet ahead from the end of the same side proportion ahead, same side video clip.

60 Variable Long Name Variable Type Counting Criteria Measurement Protocol

at least 10 square feet and within 50 Count number of large landscaping beds with shrubs or flowers that are more than 10 square large planters without trees - same side count feet, same side feet and within 50 feet from the camera.

less than 10 Count number of small planting pots with shrubs or flowers that are less than 10 square feet square feet and and within 50 feet from the camera. Small planters should be permanent elements of the within 50 feet, streetscape and not pots that are taken in at the end of the day. Do not count small planters small planters – same side count same side that are indoors and can be seen through windows.

counted landscape Rate the condition of counted landscape elements on a 1-5 Likert scale taking note of upkeep, landscape condition rating+B7 elements and lack of adequate landscaping.

landscape condition reasons text Explain landscape condition rating.

dummy passed or 500 feet Indicate the presence of common tree spacing and type on same side of street. Common tree common tree spacing and type – same side 1 = yes 0 = no ahead spacing occurs when the spacing of trees varies no more than 30%.

Indicate whether the tree spacing and type on the opposite side of the street is common to the dummy passed or 500 feet same side of the street. Common tree spacing occurs when the spacing of trees varies no common tree spacing and type – both sides 1 = yes 0 = no ahead more than 30%.

within 50 feet, moving pedestrians – same side count same side Count moving pedestrians that are within 50 feet of the camera.

61 Variable Long Name Variable Type Counting Criteria Measurement Protocol

within 50 feet, people standing still – same side count same side Count people standing still that are within 50 feet of the camera.

within 50 feet, people seated – same side count same side Count people seated that are within 50 feet of the camera.

likert scale 1 very quiet, Estimate the noise level taking note of noise from traffic, pedestrians, and any other ambient noise level 5 very loud noises.

noise level explanation text Explain noise level rating.

Count the number of distinct places that provide outdoor dining. Count outdoor dining areas dummy within 50 feet, even if there are no diners. However do not count outdoor dining if the dining area appears outdoor dining – same side 1 = yes 0 = no same side closed (umbrellas folded up, chairs on tables).

within 50 feet, tables – same side count same side Count outdoor dining tables as well as other tables that are within 50 feet from the camera.

Count number of seats that are within 50 feet from the camera. Seats around private dining do within 50 feet, not count as seats. Only count public seating. Factor in seating on planters, walls, bus stops, seats – same side count same side etc…

62 Variable Long Name Variable Type Counting Criteria Measurement Protocol

types of seating text Explain the number of seats counted. State how many seats come from planters, walls, etc…

Count pedestrian scale street lights that are at least 20% of screen height. Pedestrian scale 20% screen street lights are no more than 20 feet in height. They are oriented toward the pedestrian and pedestrian scale street lights – both sides count height, both sides are ornamented.

within 50 feet, Count other pieces of street furniture within 50 feet from the camera. Count parking meters, other street furniture – same side count same side trash cans, newspaper boxes, mail boxes, bike racks, bollards, other street lights, etc…

types of other street furniture - same side text List other street furniture counted.

Count miscellaneous street items that are within 50 feet from the camera. Count hydrants, within 50 feet, flags, banners, merchandise stands, street vendors, atms, hanging plants, flower pots, miscellaneous street items - same side count same side umbrellas, etc…

types of miscellaneous street items - same side text List miscellaneous street items counted.

20% screen height, both sides; or within 50 feet, public art - both sides count same side Count pieces of public art (sculptures, murals, etc..) that occupy at least 20% of screen height.

63 Variable Long Name Variable Type Counting Criteria Measurement Protocol

within 50 feet, Count street signs that are warning, regulatory, or directional for automobiles that are within traffic signs – same side or median count same side 50 feet from the camera.

within 50 feet, Count freestanding, hanging, and wall mounted signs that are outside buildings and are within place/building/business signs – same side count same side 50 feet from the camera. Lettering on buildings that is legible will count as building signs.

within 50 feet, directional signage - same side count same side Count directional signage oriented to the pedestrian that are within 50 feet from the camera.

20% screen height, both sides; or within 50 feet, Count billboards that occupy at least 20% of screen height or are within 50 feet from the billboards count same side camera.

dummy Indicate the presence of common signage. Common signage occurs when counted signs common signage 1 = yes 0 = no counted signs appear to have the same design.

20% screen height, both sides dummy or within 50 feet, graffiti 1 = yes 0 = no same side Indicate the presence of graffiti.

Pause video at initial view down street. Estimate proportion of screen that is sky. Estimate view sky ahead proportion proporitions in increments of 0.05

64 Variable Long Name Variable Type Counting Criteria Measurement Protocol

Pause video at initial view down street. Estimate proportion of screen that is buildings. view buildings ahead proportion Estimate proportions in increments of 0.05

Pause video at initial view down street. Estimate proportion of screen that is pavement. view pavement ahead proportion Estimate proporitions in increments of 0.05

Pause video at initial view down street. Estimate proportion of sreen that is cars. Estimate view cars ahead proportion proporitions in increments of 0.05

Pause video at initial view down street. Estimate proporiton of screen that is street furniture. Pedestrian scale street lights, tables, seating, other street furniture, other miscellaneous street view street furniture ahead proportion items and public art all count as street furniture. Estimate proporitions in increments of 0.05

Pause video at initial view down street. Estimate proportion of screen that is landscaping. view landscaping ahead proportion Estimate proporitions in increments of 0.05

Pause video at initial view down street. Estimate proportion of screen that is occupied by elements that cannot be categorized as sky, buildings, pavement, cars, street furniture, or landscaping (for example, people, free standing signs). Estimate proportions in increments of view other ahead proportion 0.05.

types of other ahead text List elements that were categorized as other ahead

65 Variable Long Name Variable Type Counting Criteria Measurement Protocol

Pause video at intital view across street. Estimate proportion of screen that is sky. Estimate view sky across proportion proporitions in increments of 0.05

Pause video at initial view across street. Estimate proportion of screen that is buildings. view buildings across proportion Estimate proporitions in increments of 0.05

Pause video at initial view across street. Estimate proportion of screen that is pavement. view pavement across proportion Estimate proporitions in increments of 0.05

Pause video at initial view across street. Estimate proportion of screen that is cars. Estimate view cars across proportion proporitions in increments of 0.05

Pause video at initial view across street. Estimate proporiton of screen that is street furniture. Pedestrian scale street lights, tables, seating, other street furniture, other miscellaneous street view street furniture across proportion items and public art all count as street furniture. Estimate proporitions in increments of 0.05

Pause video at initial view across street. Estimate proportion of screen that is landscaping. view landscaping across proportion Estimate proporitions in increments of 0.05

Pause video at initial view across street. Estimate proportion of screen that is occupied by elements that cannot be categorized as sky, buildings, pavement, cars, street furniture, or landscaping (for example, people, free standing signs). Estimate proportions in increments of view other across proportion 0.05.

66 Variable Long Name Variable Type Counting Criteria Measurement Protocol

types of other across text List elements that were categorized as initial othe across.

maximum value of view ahead proportion Record the largest value of the view ahead variables

maximum value of view across proportion Record the largest value of the view across variables

67 Appendix 4: Relationships Between Urban Design Qualities and Physical Features (hypothesized in regular type and validated in bold)

Variable Imageability Legibility Enclosure Scale Human Transparency Linkage Complexity Coherence Tidiness courtyards/plazas/parks - both sides X X X X arcade – same side X X X X landmarks - both sides X X major landscape features - both sides X X memorable architecture X X distinctive signage X X long sight lines X X X X terminated vista X X X X X progress toward next intersection X X proportion of distance walked versus distance visible X X street connections to elsewhere X X number of buildings – both sides X number of land uses – both sides X proportion of historic building frontage - both sides X number of buildings with identifiers - both sides X X proportion of building frontage with identifiers various building ages X X number of primary building materials X X number of dominant building colors X X X number of accent colors - both sides X X building projections – same side X X visible sets of doors – same side X X X X visible recessed doors – same side

68 Variable Imageability Legibility Enclosure Scale Human Transparency Linkage Complexity Coherence Tidiness proportion of counted sets of doors that are recessed X proportion first floor facade with windows X X proportion entire facade with windows X X common window proportions - both sides X awnings or overhangs – both sides X X X X height interruptions – same side X number of buildings with non-rectangular silhouettes X X proportion of buildings with non-rectangular silhouettes X common architectural style - both sides X X common materials – both sides X proportion active uses – same side X X X proportion street wall – same side X X proportion street wall – opposite side X enclosed sides X average building setback X common setbacks – both sides X X building height – same side X X X building height to width ratio – same side X building height – opposite side X X common building heights – both sides X X X common building masses – both sides X street width X X X median width X sidewalk width - same side X building height to street width ratio X

69 Variable Imageability Legibility Enclosure Scale Human Transparency Linkage Complexity Coherence Tidiness sidewalk clear width - same side X buffer width - same side X number of paving materials X X textured sidewalk – same side X X X textured street X X X pavement condition X debris condition X parked cars – same side X proportion of street with parked cars – same side X X moving cars – both sides X X X speed X traffic to street width ratio X moving cyclists - both sides X X curb extensions – same side X X X midblock crossings X X X midblock passageways – same side X X X X overhead utilities X X number of landscape elements - both sides X landscaped median X X X number of trees – both sides X X X trees in wells or landscaped beds - same side X X proportion of sidewalk shaded by trees – same side X large planters without trees - same side X X small planters – same side X X landscape condition X

70 Variable Imageability Legibility Enclosure Scale Human Transparency Linkage Complexity Coherence Tidiness common tree spacing and type – same side X X X common tree spacing and type – both sides X X X X pedestrians moving – same side X X X X stationary people standing – same side X X X people seated – same side X X X X noise level X X X outdoor dining – same side X X X X X tables – same side X X seats – same side X X pedestrian scale street lights – both sides X X X X other street furniture – same side X X X X miscellaneous street items - same side X X X public art - both sides X X X X traffic signs – same side or median X X X place/building/business signs – same side X X X X directional signage - same side X X billboards X X X X common signage X X graffiti X X sky ahead X buildings ahead X pavement ahead cars ahead street furniture ahead X landscaping ahead X X

71 Variable Imageability Legibility Enclosure Scale Human Transparency Linkage Complexity Coherence Tidiness sky across X buildings across X pavement across cars across street furniture across X landscaping across X X maximum ahead X maximum across X

72 Appendix 5: References Alexander, C., S. Ishikawa, and M. Silverstein. 1977. A Pattern Language - Towns Buildings Construction. Oxford University Press, New York.

Christopher Alexander. 1965 (April). A City is Not a Tree. Architectural Forum. 122 (No 1&2): 58-62.

Appleyard, D. 1981. Livable Streets. University of California Press, Berkeley, CA.

Arnheim, R. 1977. Dynamics of Architectural Form. University of California Press, Berkeley, CA.

Arnold, H. 1993. Trees in Urban Design. Van Nostrand Reinhold, New York.

Bacon, E. 1974. Design of Cities, Viking Press, New York.

Blumenfeld, H. 1953 (April). Scale in Civic Design, Town Planning Review. 24: 35-46.

Cullen, G. 1961. The Concise Townscape. Reed Educational and Professional Publishing, London.

Duany, A. and E. Plater-Zyberk. 1992. The Second Coming of the American Small Town. Wilson Quarterly. 16: 19-48

Elshestaway, Y. 1997. Urban Complexity: Toward the measurement of the physical complexity of streetscapes. Journal of Architectural and Planning Research. 14: 301-316.

Ewing, R. 1996. Pedestrian- and Transit-Friendly Design. Tallahaseee, FL.

Ewing, R. 2000. Asking Transit Users about Transit-Oriented Design. Transportation Research Record. 1735: 19-24.

Ewing, R. 2005. Can the Physical Environment Determine Physical Activity Levels? Exercise and Sport Sciences Review, pages not assigned.

Ewing, R. (in press) Turning Highways into Main Streets: Two Innovations in Planning Methodology. Journal of the American Planning Association.

Ewing, R. and R. Cervero. 2001. Travel and the Built Environment. Transportation Research Record 1780: 87-114.

73

Gehl, J. 1987. Life Between Buildings–Using Public Space, Van Nostand Reinhold, New York.

Handy, S. 1992. Regional versus Local Accessibility: Variations in Suburban Form and the Effects on Nonwork Travel. Unpublished dissertation, University of California Berkeley.

Handy, S. 2004. Critical Assessment of the Literature on the Relationships Among Transportation, Land Use, and Physical Activity.

Heath, T, S. Smith, and B. Lim. 2000. The Complexity of Tall Building Facades. Journal of Architectural and Planning Research. 17(3): 206-220.

Hedman, R. 1984. Fundamentals of Urban Design, American Planning Association, Chicago, IL.

Herzog, T.R., S. Kaplan, and R. Kaplan. 1982. The Prediction of Preference for Unfamiliar Urban Places. Population and Environment. 5(1): 43-59.

Herzog, T.R. and O.L. Leverich. 2003. Searching for Legibility. Environment and Behavior. 35(4): 459-477.

Jacobs, A. 1993. Great Streets. MIT Press, Cambridge, MA.

Jacobs, A. and D. Appleyard. 1987. Toward an Urban Design Manifesto. Journal of the American Planning Association. 53: 112- 120.

Jacobs, J. 1961. The Death and Life of Great American Cities. Random House, New York.

Kaplan, R. and S. Kaplan. 1989. The Experience of Nature—A Psychological Perspective. Cambridge University Press, New York.

Kay, J.H. (1997) Asphalt Nation: How the Automobile Took over America, and How We Can Take It Back. University of California Press, Berkeley.

Landis, J.R. and G.G. Koch. 1977. The Measurement of Observer Agreement for Categorical Data. Biometrics. 33:159-174.

Lennard, S.H.C. and H.L. Lennard, Livable Cities—People and Places: Social and Design Principles for the Future of the City, Center for Urban Well-Being, Southhampton, New York, 1987.

74

Lynch, K. 1960. The Image of the City. Joint Center for Urban Studies, Cambridge, MA.

Nasar, J.L. 1987. The effect of sign complexity and coherence on the perceived quality of retail scenes. Journal of the American Planning Association. 53: 499-509.

Nelessen, A. 1994. Visions for a New American Dream. American Planning Association, Washington, D.C.

Rapoport, A. 1990. History and Precedent in Environmental Design. Kluwer Academic Publishers, Plenum Press, New York.

Raudenbush S.W., A. Bryk, and R. Congdon. 2004. HLM 6: Hierarchical Linear & Nonlinear Modeling. Scientific Software International, Chicago, IL.

Sitte, C. 1889. City Planning According to Artistic Principles. Verlag von Carl Graeser, Vienna (complete translation in G.R. Collins, C.C. Collins, Camillo Sitte. 1986. The Birth of Modern City Planning. Rizzoli International Publications, New York).

Stamps, A.E. 1998. Measures of Architectural Mass: From Vague Impressions to Definite Design Features. Environment and Planning B. 25: 825-836.

Stamps, A.E. 1999. Sex, Complexity, and Preferences for Residential Facades. Perceptual and Motor Skills. 88: 1301-1312.

Trancik, R. 1986. Finding Lost Space—Theories of Urban Design. Van Nostrand Reinhold, New York.

Tunnard, C. and B. Pushkarev. 1963. Man-Made America—Chaos or Control? Yale University Press, New Haven, CT.

Unwin, R. 1909. Town Planning in Practice. T. Fisher Unwin, Ltd., London (reissued by Princeton Architectural Press: New York, 1994).

Whyte, W.H. 1988. City—Rediscovering the Center. Doubleday, New York.

75