ca rtographicathe international journal for geographic information and geovisualization

cartographica (volume 44, issue 3), pp. iii doi:10.3138/carto.44.3.iii iii

Cartographica / Contents

Volume 44, Number 3, Fall 2009

Special Issue: Cognitive Issues in Geographic Information VisualizationGuest Editors' Introduction

Sara Irina Fabrikant and Amy LobbenApplications of Signal Detection Theory to Geographic Information Science

Amy L. Griffin and Scott BellfMRI and Human Subjects Research in Cartography

Amy Lobben, Megan Lawrence, and Judy M. OlsonNaïve Cartography: How Intuitions about Display Configuration Can Hurt Performance

Mary Hegarty, Harvey S. Smallman, Andrew T. Stull, and Matt S. CanhamNavigation Tasks with Small-Display Maps: The Sum of the Parts Does Not Equal the Whole

Julie A. DillemuthIssues of Change Detection in Animated Choropleth Maps

Kirk Goldsberry and Sarah BattersbyColour-Enhanced Star Plot Glyphs: Can Salient Shape Characteristics Be Overcome?

Alexander Klippel, Frank Hardisty, Rui Li, and Chris Weaver

/ 139

/ 145

/ 159

/ 171

/ 187

/ 201

/ 217

Advice to ContributorsThe Canadian Cartographie Association / L'Association cartographique canadienneThe International Cartographie Association / L'Association cartographique internationale

/ 233

/ 235

/ 236

------------------------------~-~-~-~~~~~~-~-~-,~~~~~~,~,~~~~--....------..-----------------

Introduction: Cognitive Issues in GeographieInformation Visualization

Sara Irina FabrikantDepartment of Geography / University of Zurich / Zurich / Switzerland

Amy LobbenDepartment of Geography / University of Oregon / Eugene / OR / USA

In recent years geovisualization has morphed into thefield of "geovisual analytics," emphasizing humanexploration through highly interactive visual interfacesto geographic information, taking advantage of perceptualand cognitive abilities to recognize and process patternsand outliers from visual displays, linking these patternsand outliers to existing mental schemata and knowledgebases, and eventually arriving at an appropriate spatiallyrelevant course of action or decision given the visual input(Chen and others 2008). Visual analytics is based on theintuition that highly interactive and dynamic depictionsof complex and multivariate databases amplify humancapabilities for inference and decision making, as theyfacilitate cognitive tasks such as pattern recognition, ima-gination, association, and analytical reasoning (Thomasand Cook 2005). One may wonder, however, why successstories in this vibrant (albeit still young) research fieldseem practically non -existent (Johnson and others2006). By «success story" we mean a scenario in whichthe creators of a visual analytics tool or approach are notthe sole user group, or significant insights into real-worlddata were generated for decision makers, or target userscould actually use the tool to solve real research problemsor perform the day-ta-day tasks at hand.

Unsurprisingly, one pervasive theme underlying manycurrent (geo- )visualization research challenges is the dif-ficulty of effectively evaluating highly interactive visuali-zation tools and complex displays and identifying theirpotentially positive influence on exploratory data analysis,knowledge extraction, and learning. While research within

the visual analytics and geovisual analytics communitieshas focused predominantly on building tools and thehighly interactive computer-human interfaces thataccompany them, fundamental investigations based onempirical evidence of how, when, and why human infer-ence, analytical reasoning, and decision makingwith visual displays work have received considerably lessattention. We still know little about the effectivenessof graphical displays for exploratory data analysis,problem solving, knowledge exploration, learning,and decision making. This lack of understanding is parti-cularly noteworthy because the predecessor of thecurrent International Cartographic Association (ICA)Commission on Geovisualization (then called the ICACommission on Visualization and VirtualEnvironments) had already identified this problem atthe beginning of the twenty-first century as one of thekey challenges for geovisualization research - namely,the necessity to focus on cognitive issues and usability(MacEachren and Kraak 2001). Specifically, these authorspoint to the need to develop a theoretical framework,based on cognitive principles, to support a human-centred approach to geovisualization. In a 2006 reporton visualization research challenges that not only sum-marizes visualization achievements but also discussesmajor obstacles blocking the discipline's advancement,Johnson and others (2006, 9-10) emphasize that if weare to accelerate progress in visualization and visual com-munication' the investigation of the nature, options, lim-itations, and effects of human perception, humancognition, and the visual exploration experience needs

cartographica (volume 44, issue 3), pp. 139-143 doi:10.3138/carto.44.3.139 139

Sara Irina Fabrikand and Amy Labben

to be addressed more systematically. Along these lines,Gennady Andrienko and others (2008) point out thatvisual analytics methods developed must be cognitivelyadequate for successful inference and that tools must beusable for analysis and decision making. One appropriatestarting point for constructing effective and efficientvisual analytics tools is to frame developments withincontext of cognitive theories, long-standing empiricallyevaluated design principles, and related empirical studieson visual displays.

This challenge inspired the idea for this special issue ofCartographica. The aim of the issue is to showcase currentfundamental empirical research and state-of-the-art eva-luation methods within the interdisciplinary empiricalresearch domain of geographic information visualizationand cognition, as well as to further stimulate new ideas forfuture empirical work. The papers originated from con-tributions presented in two paper sessions at theAssociation of American Geographers (AAG) annualmeeting in Boston, Massachusetts, on 18 April 2008.These sessions,' sponsored by the Cartography, GIS, andBehavioral Geography and Environmental PerceptionSpecialty Groups of the AAG, were the result of an inter-nationally disseminated open call for papers initiated bythe guest editors of this issue. The call for papers specifi-cally aimed to attract theoretical and empirical contribu-tians exploring the human-geovisualization displayinterface. Topics of interest listed in the call for papersincluded

• geovisualization design research (2- and 3D, static,animated and interactive, virtual and immersive,mobile, etc.)

• the application of cognitive theories and methodsto understanding visuo-spatial displays and tooluse for inference and decision making

• the application of visuo-spatial displays and tools tounderstanding spatial cognition

• spatial reasoning, inference, and decision makingwith visuo-spatial displays and tools

• human-visualization interaction research

The high level of interest in cognitive work raised by thecall for papers ultimately resulted in the participation ofnine presenters based at universities all over the UnitedStates, as well as one from Switzerland, and attracted anaudience of more than 50 people for each session. Theenthusiastic response to the AAG sessions, from pres en -ters and the audience alike, and the appearance of thisspecial issue suggest that the interdisciplinary topic ofcognition in geovisualization is not only timely and rele-vant for the GIScience, geovisualization, and cartographyresearch communities but also of significant interest toa broader interdisciplinary audience, including spatialcognition researchers. Geovisualization research seemsalive and well, considering the growing numbers of

140

workshops, meetings, and conferences, for example,organized by the ICA Commission on GeoVisualisation/or its active members and the resulting papers in proceed-ings' journals, and special journal issues in GIScience,cartography, and other cognate fields.

After the AAG conference, the session organizers invitedpresenters to submit full papers for review and potentialpublication in Cartographica; this invitation was followedby a second open call for papers, disseminated digitallythrough various e-mail lists around the globe. The sevenfull papers received included four submissions extendedfrom oral AAG presentations and three external submis-sions responding to the second call for papers. Eachsubmission was double-blind reviewed by two to threeexternal reviewers as well as by the journal editorand the guest editors. The journal editor handled theco-authored contribution by one of the guest editorsseparately. The six papers that appear in this specialissue are the result of this multi-stage review process.

While the original topics of interest proposed in the callsfor papers included more specific foci, we collapsed thetopics into two conference sessions titled "Methods" and"Domains," while recognizing that some papers fit neatlyinto both sessions. This special issue features articles thatare unique in offering both novel methods for empiricalwork in geovisualization and new domains for applyingthese methods. Because a rigid categorical breakdowndoes a disservice to these materials, we will not presumeto pigeonhole each article into one of the specificfoci mentioned above. In these articles, new methodsfor the geovis community are presented in domain con-texts (Griffin and Bell; Lobben, Lawrence, and Olson);new domains are presented with established methodsfrom other research disciplines (Hegarty and others;Dillemuth); and hybrid papers address current cartog-raphic constructs in new ways and apply methods fromother research areas to answer fundamental cartographicquestions (Goldsberry and Battersby; Klippel and others).

Two studies propose knowledge transfer from establishedevaluation methods in psychology to solve fundamentalbehavioural research questions in geography using visualdisplays. Amy L. Griffin and Scott Bell suggest the appli-cation of signal detection theory (SDT), while AmyLobben and others propose neuroimaging, and specifi-cally functional magnetic resonance imagine (MRI), tofurther expand the cartographic research methodology.Griffin and Bell suggest that SDT offers cartography,GIScience, and visualization an analytical method withthe potential to enhance unique properties associatedwith much past and current (and, likely, future) research.They suggest that SDT is potentially applicable in researchwhose objectives include decision making involving mul-tiple choices, as, they argue, does much of our research.Griffin and Bell characterize SDT as allowing GIScienceresearchers to posit questions that lead to four potential

cartographica (volume 44, issue 3)

responses: a hit (correctly identifying an object), a falsealarm (identifying a false object), a miss (failing to iden-tify the correct object), and correct rejection (correctlyrejecting the incorrect object). In arguing for the useful-ness and effects of SDT, they use the example of a humannavigation study in which volunteers were asked to followa route depicted on one of three navigation aids: a surveymap, a strip map, and a segmented strip map. Half theparticipants studied the map prior to navigation andtraced the route from memory, while half traced theroute while carrying the map for reference. The authors'application of SDT allowed them to discover an effect ofnavigating from memory versus navigating with a map:participants who travelled with the map were less likely toremember landmarks. Griffin and Bell suggest that such adiscovery would likely not have been possible without theapplication of SDT. In addition to the case study, theysuggest several potential applications of SDT in geographyand provide general examples. They further suggest thatSDT has the potential to be useful in research thatincludes non-human (i.e., machine) decision making.

The paper by Lobben and others is the first fMR_! reviewstudy to appear in the peer-reviewed GIScience literature.Their study provides a potential foundation for the inclu-sion of neuroimaging methods - specifically fMRI - incartographic research, situating their argument withinwell-known research areas in cartographic design: visualsearch, perceptual grouping, and figure-ground. Lobbenand others review studies from the neuroscience literaturethat are similar (in terms of terminology and concepts, ifnot of stimuli) to these cartographic design constructs.The objective of the paper is not to suggest activationassociated with cartographic design (as the authorspoint out, such studies have yet to be conducted) but tosuggest a framework for the development of fMRI-relatedcartographic design research.

New domains are proposed and investigated by cognitivepsychologists Mary Hegarty and others and by a geogra-pher, Julie A. Dillemuth, who approaches a classic empiri-cal research area in behavioural geography, navigation,from a new perspective. Hegarty and others proposeand investigate "naïve cartography," which they suggestrefers to individuals' intuitions about visual displays andtheir design choices and opinions. They document aninteresting dichotomy: naïve cartographers tend toprefer 3D over 2D representations and to prefer morerealistic (and more complex) representations to simpler,more streamlined representations of the same construct.But principles of cartographic and graphic design suggestthe opposite: abstraction, simplification, and generaliza-tion. The authors conducted a study to ascertain whethernaïve cartographers distinguish between display prefer-ences and display efficiencies. They also investigated theeffect of experience on naïve cartographic intuition. In thefirst of three studies, they disseminated a survey to more

cartographica (volume 44, issue 3)

Introduction: Cognitive Issues in Geographic Information Visualization

than 700 undergraduate students (a group they define asnaïve cartographers); the respondents preferred enhanceddisplays with animation and realism. For the secondstudy, Hegarty and others altered their questionnaireand selected meteorologists - who regularly work withvisual displays - as an "expert" group. The resultsrevealed similar a preference for enhanced displays withanimation and realism. The authors further investigatedthe relationship between preference and performance ofmap tasks among naïve undergraduates and meteorolo-gists. Their experimental design included a spatial abilitiestask and tasks requiring participants both to make infer-ences and to rate preference between maps that varied interms of realism. The results of this third study revealedthat enhanced realism on weather maps led to declines inperformance for both groups but that participants stillpreferred the enhanced maps. Hegarty and others' overallresults suggest a dichotomy between preference and effec-tiveness: map users' preference for a display type is notnecessarily a good indication of the effectiveness of thatdisplay type ..These results are another good, empiricallyvalidated illustration of the common -sense notion thatwhat people think they want is not always what is bestfor them. Hegarty and others offer a new cartographicconstruct, naïve cartography, along with some interestingsupporting empirical data.

While previous research has been conducted on map useand navigation, Dillemuth provides a new perspectivethrough her investigation of the effects of small digitalmobile map displays (i.e., maps on cell phones or personaldigital assistants) on map use, navigation, and cognitivemap development. Her volunteers, grouped into one offour conditions (100/0, 250/0, 400/0, or 1000/0 map view),were asked to perform the same tasks: distance/directionestimates, landmark recall, and sketch mapping. All par-ticipants also completed a sense-of-direction question-naire as well as a perspective taking/spatial orientationtest. Overall, as map extent decreased (from 1000/0 to100/0), accuracy declined and response times rose.Though no difference in confidence was observed betweengroup conditions, Dillemuth found that higher spatialabilities (as measured by the instruments used in- thestudy) correlated with higher performance in the 100/0viewing condition.

Finally, two papers provide an effective hybrid approach.Both Kirk Goldsberry and Sarah Battersby and AlexanderKlippel and others address familiar cartographic topics:map animations and multivariate map symbols (i.e.,glyphs). In both cases, however, the authors add a newwrinkle by using methods borrowed from other researchfields to evaluate and understand users' perception of andunderstanding from visual displays. Goldsberry andBattersby focus on enhancing the understanding of ani-mated choropleth maps; in doing so, they offer new meth-odologies for investigating the misunderstanding of these

141

Sara Irina Fabrikand and Amy Labben

maps. Their study systematically explores why animatedmaps are difficult to read. In addition to a review ofrelevant literature in the areas of change blindness,foveal and peripheral attention, smooth and abruptchanges, visual variables, spatial attention, and magnitudeof change, the authors address potential causes of misun-derstanding or misreading of animated choropleth maps,including the limits of the human perceptual system andthe limits of the research methodologies currentlyemployed to understand these causes. Moreover, theyprovide a foundation for launching a systematic metho-dological approach to empirical research designedto investigate the difficulties associated with animatedchoropleth maps. This foundation includes methods forquantifying change in animated choropleth maps basedon detecting and quantifying magnitude of change.Goldsberry and Battersby suggest that magnitude ofchange is a critical confounding factor in understandinganimated choropleth maps, and propose two methodsfor quantifying magnitude of change. Basic magnitudeof change (BMOC) is a count of the enumeration unitsin the animated choropleth map that change between theorigin -state and the destination -state (or the beginningand ending of the animation); the proportion of changeis then normalized by dividing the changed units by totalnumber of units. Magnitude of rank change (MORC)provides an indication of the severity of change betweenthe origin-state and the destination-state of an enumera-tion unit.

Klippel and others also tackle a fundamental cartographicdesign issue: the influence of perception on the extractionof meaning from map symbols. Following on fromtheir previous research, they conducted two experimentson shape symbolization, including perceptual classifica-tion, and categorization. Eight-point star-plot glyphswere used to represent data about future cars, with eachvariable (price, weight, safety rating, maximum speed,miles per gallon, acceleration, emissions, and interiorspace) represented by one of the eight rays. Because thefindings of a previous experiment revealed that single-spike glyphs resulted in unique grouping properties,Klippel and others introduced colour as a variable, in anattempt to offset the influence of shape. Their study alsoincluded a control group that viewed varying glyph shapesnot related to the data set (geometric shape comparisonsonly). Their volunteers were assigned to one of two groups.All participant groups viewed 81 glyphs; eight variables/rays received four different treatments in the glyph designs.Four of the variables were assigned values categorizedinto three classes (low, medium, high); the other fourwere assigned randomized values within the "high" cate-gory. Participants were asked to view the glyphs and createa categorization; they were then asked to explain theirgrouping rationale. The empirical results, not surprisingly,reveal that adding meaning to the glyphs (in this case,

142

values for the variables/rays) slowed decision making(measured by reaction time or time to finish) relative toshape-only classification. Like Goldsberry and Battersby,Klippel and others also introduce a new analyticalmethod, KlipArt, designed and applied to analysis of dif-ferences in icon classification. Their discussion suggeststhat future research should include an evaluation ofpoint-symbol properties and their effects on pre-attentiveprocessing to facilitate visual search, perceptual grouping/categorization, and decision making.

This point brings this special issue near full circle. Lobbenand others also address methods to investigate perceptualgrouping; what if Klippel and others applied MRI toinvestigate the brain-activation differences between thefour glyph conditions? Perhaps they could hypothesizethe observation of different activation patterns associatedwith each of the four conditions.

With this collection of six papers, we hope to have raisedinterest in cognitive issues in geovisualization and to sti-mulate new research avenues toward more systematicinvestigations of the effects of human perception and cog-nition with highly interactive, multivariate, multidimen-sional, and dynamic geovisualization displays.

Acknowledgements

This special issue would not have been possible withoutthe commitment, perseverance, and help of many people.We would like to thank Cartographica Editor JeremyCrampton, as well as Co-editors Clifford Wood andRoger Wheate and Editorial Assistant Laura Wilson,who were an invaluable help in making this special issuehappen. Special thanks go to the reviewers for their excel-lent support under considerable time constraints. Last butnot least, we thank all the authors who took the time andeffort to submit their best work to the 2008 AAG sessionsin Boston and to this special issue.

Author Information

Sara Irina Fabrikant, Associate Professor of Geographyand Head of the Geographic Information Visualizationand Analysis Unit, Department of Geography,University of Zurich, Winterthurerstr. 190, CH-8057Zurich, Switzerland. Tel: +41 (44) 635-5150. Fax: +41(44) 635-6848. E-mail: sara@geo.uzh.ch. ProfessorFabrikant's research and teaching interests lie in geo-graphic information visualization and (geo )visual analy-tics, GIScience and cognition, and graphical user interfacedesign and evaluation, including dynamic cartography,

Amy Lobben, Assistant Professor, Department ofGeography, 1251 University of Oregon, Eugene, OR97403 USA. Her research and teaching interests lie inenvironmental and representational perception, mapuse, navigation, spatial and environmental cognition,

cartographica (volume 44, issue 3)

and geographic/cartographic applications In neuroima-ging (specifically fMRI).

Notes

1. The sessions are available online at http://communicate.aag.org/eseries/aag_org/prog ram/Session Deta il.cfm?SessionlD=5261 (session 1) and http://communicate.aag.org/eseries/aag_org/prog ram/Session Deta i I.cfm ?SessionID=6181 (session 2)

2. The homepage of the ICA Commission on GeoVisualizationcan be accessed at http://geoanalytics.net/ica/.

References

Andrienka, G., N. Andrienka, J. Dykes, 5.1. Fabrikant, andM. Wachowicz. 2008. "Geovisualization of Dynamics,Movement and Change: Key Issues and Developing

cartographica (volume 44, issue 3)

Introduction: Cognitive Issues in Geographic Information Visualization

Approaches in Visualization Research:' InformationVisualization 7/3: 173-80.

Chen, J, R. Roth, A. Naito, E. Lengerich, and A. MacEachren.2008. "Geovisual Analytics to Enhance Spatial ScanStatistic Interpretation: An Analysis of U.S. Cervical CancerMortality:' International Journal of Health Geographies 7: 57.Available at http://www.ij-healthgeographics.com/content/7/l/57.

Johnson, C.R., R. Moorhead, T. Munzner, H. Pfister,P. Rheingans, and T.S. Yoo. 2006. NIH-NSF VisualizationResearch Challenges Report Los Alamitos, CA. IEEEPress.

MacEachren, A.M., and M.-J. Kraak. 2001. "ResearchChallenges in Geovisualization" Cartography andGeographic Information Science 28: 3-12.

Thomas, J.J., and K.A. Cook. 2005. Illuminating the Path:Research and Development Agenda for Visual Analytics.Richland, WA: IEEE Press.

143

c a rto graph ic athe international journal for geographic information and geovisualization

"a