V ISUALIZATION C ORNER

Editors: Jim X. Chen, [email protected] R. Bowen Loftin, [email protected]

VISUALIZATION RESEARCH CHALLENGES A Report Summary

By Robert Moorhead, Chris Johnson, , Hanspeter Pfister, Penny Rheingans, and Terry S. Yoo

EARLY 20 YEARS AGO, THE US NATIONAL SCIENCE FOUN- explains how enables ex- periments of complex phenomena. Vi- N DATION (NSF) CONVENED A PANEL TO REPORT ON VI- sualization reduces and refines data streams, letting us winnow huge vol- SUALIZATION’S POTENTIAL AS A NEW TECHNOLOGY.1 IN 2004, umes of data in applications—for ex- ample, public health surveillance at a THE NSF AND THE US NATIONAL INSTITUTES OF HEALTH (NIH) regional or national level that tracks the spread of infectious diseases. The convened the Visualization Research Visualization’s Value “Virus Structures” sidebar shows how Challenges (VRC) Executive Com- Advances in computing science and visualization helps structural biologists mittee—made up of this article’s au- technology have engendered unprece- understand how virus structure corre- thors—to write a new report. Here, we dented improvements in scientific, bio- lates with strength. Visualizations of summarize that report, available in full medical, and engineering research; such application problems as hurri- at http://tab.computer.org/vgtc/vrc/ defense and national security; and in- cane dynamics and biomedical imag- and in print.2 dustrial innovation. Continuing and ing are generating new knowledge that The VRC report aims to accelerating these advancements will crosses traditional disciplinary bound- require comprehending vast amounts aries. Visualization can also provide • evaluate the progress of the matur- of data and information produced from industry with a competitive edge by ing visualization field, myriad sources.3 Visualization—namely, transforming business and engineer- • help focus and direct future research helping people explore or explain data ing practices into more understand- projects, and through software systems that provide able procedures. • provide guidance on how to appor- a static or interactive visual represen- Although well-designed visualiza- tion national resources as research tation—is critical to achieving this tions can help people enormously, challenges change rapidly in the fast- goal. Visualization designers can ex- naive visualization attempts are all too paced information technology world. ploit the high-bandwidth channel of often ineffective or even actively mis- human and help us leading. Designing effective visualiza- We explore the state of the field, exam- comprehend information orders of tions is a complex process that requires ine visualization’s potential impact on magnitude more quickly than we a sophisticated understanding of hu- areas of national and international im- would reading raw numbers or text. man information-processing capabili- portance, and present our findings and Visualization is fundamental to un- ties, both visual and cognitive, and a recommendations for the future of this derstanding models of complex phe- solid grounding in the considerable growing discipline. Our audience is nomena, such as multilevel models of body of work that already exists in the twofold: visualization research’s sup- human physiology from DNA to visualization field. Further research in porters, sponsors, and application users whole organs, multicentury climate visualization—and the transfer of ef- on the one hand, and visualization re- shifts, international financial markets, fective visualization methodologies searchers and practitioners on the other. or multidimensional simulations of into the working practice of medicine, Our findings and recommendations re- airflow past a jet wing. The “Charac- science, engineering, and business— flect information gathered from visual- terizing Methods” will be critical in handling the ongoing ization and applications scientists during sidebar summarizes a study of several information explosion. two VRC workshops, as well as input flow visualization techniques, whereas Although visualization is itself a dis- from the larger visualization community. the “Fusion Plasma Physics” sidebar cipline, advances in visualization lead

66 Copublished by the IEEE CS and the AIP 1521-9615/06/$20.00 © 2006 IEEE COMPUTING IN SCIENCE & ENGINEERING CHARACTERIZING FLOW overall: on average, subjects were fastest and most accurate ISUALIZATION ETHODS when using it. V M This study produced both quantitative results and a basis or decades, researchers have been developing and pub- for comparing other visualization methods, creating more F lishing visualization techniques that advance the state of effective methods, and defining additional tasks to further the art. However, disproportionately few quantitative stud- understand trade-offs among methods. A future challenge is ies compare visualization techniques. Figure A shows the to develop evaluation methods for more complex 3D time- differences between flow visualization methods, showing varying flows. six methods for visualizing the same 2D vector field.1 Sub- jects who participated in the user study performed several Reference tasks, including identifying the type and location of critical 1. D.H. Laidlaw et al., “Comparing 2D Vector Field Visualization Meth- points in visualizations. Assuming roughly equal importance ods: A User Study,” IEEE Trans. Visualization and , for all tasks, the streamline visualization performed best vol. 11, no. 1, 2005, pp. 59–70.

(1) (2) (3)

(4) (5) (6)

Figure A. Six methods for visualizing the same 2D vector field. They include (1) icons on a regular grid; (2) icons on a jittered grid; (3) layering method inspired by oil painting; (4) line-integral convolution; (5) image-guided streamlines; and (6) streamlines seeded on a regular grid. (Figure courtesy of David Laidlaw, Brown University.)

inevitably to advances in other disci- curity, medicine, sociology, and public interpreting information, both quanti- plines. Just as knowledge of mathe- policy, so too is visualization becoming tatively and qualitatively, and with pre- matics and statistics has become indispensable in enabling researchers senting data in a way that most clearly indispensable in subjects as diverse as in other fields. Like statistics, visual- conveys their salient features. Both the traditional sciences, economics, se- ization is concerned with analyzing and fields develop, understand, and abstract

JULY/AUGUST 2006 67 V ISUALIZATION C ORNER

FUSION PLASMA PHYSICS

tokamak is a doughnut-shaped chamber used in fusion A research. During tokamak experimental operation, when a plasma is heated and confined magnetically, events occasionally occur that rapidly terminate the plasma dis- charge. For future experiments, such as the International Thermonuclear Experimental Reactor (ITER), the stored en- ergy will be approximately 100 times greater than in pre- sent-day devices. Thus, these disruptions1 have the potential to severely damage the material wall, especially if the heat flux is highly localized. Figure B shows the results of a simu- lation of a particular disruption in the DIII-D tokamak. The Figure B. Tokamak disruption simulation. The image figure presents visualizations of the temperature isosurfaces, presents visualizations of the temperature isosurfaces, the magnetic field lines, and contours of the heat flux on the the magnetic field lines, and contours of the heat flux wall, which show toroidal and poloidal localization as a re- on the wall. sult of the magnetic field lines’ topology. The localization re- sults from the plasma instability compressing the flux surface in the plasma’s core and then transporting the resulting “hot Reference spots” to the wall. Visualizing the data in 3D shows where 1. S.E. Kruger, D.D. Schnack, and C.R. Sovinec, “Dynamics of the Major the plasma wants to preferentially bulge, enabling the devel- Disruption of a DIII-D Plasma,” Physics of Plasmas, vol. 12, no. opment of improved disruption mitigation techniques. 056113, 2005.

data-analytic ideas and package them as Visualization Hardware computationally intensive. A great techniques, algorithms, and software Many of the original report’s hardware deal of academic and industrial re- for various application areas. concerns have been allayed over time search in software volume-rendering However, despite visualization’s im- and by Moore’s law. Processors with algorithms helped the field mature portance to discovery, security, and what used to be considered supercom- and finally made hardware creation competitiveness, support for research puter-class power are now available in feasible. Grant-funded university re- and development in this critical, multi- commodity desktop PCs that cost a few search that began at the State Univer- disciplinary field has been inadequate. thousand dollars. Graphics perfor- sity of New York (SUNY) led to the Unless we recommit ourselves to sup- mance that used to require special-pur- development of an actual product porting visualization research, devel- pose workstations costing tens or through Mitsubishi Electric Research opment, and technology transfer, we’ll hundreds of thousands of dollars is Lab (MERL), culminating in the suc- see a decline in the progress of discov- now available as a commodity graphics cessful spin-off company TeraRecon ery in the other important disciplines card for a few hundred. Fast and cheap (www.terarecon.com). that depend on visualization. As these hardware aimed at the business and en- In contrast, display technology im- disciplines lose their ability to harness tertainment mass markets allows un- provements have historically lagged and interpret information, the discov- precedented access to computational far behind the Moore’s law curve. In ery rate itself will decline. In the in- and graphics power for visualization— the past 20 years, cathode-ray-tube evitable chain reaction, the US will a boon for both visualization resear- (CRT) displays have little more than lose its competitive edge in business chers and end users. The latest doubled in physical display size and and industry. generation of programmable graphics resolution and have retained the same pipelines on these cards are flexible weight and form factor. In the past, State of the Field: enough to spark an explosion of so- our ability to design user interfaces Infrastructure phisticated processing techniques that has been constrained because moni- One of the 1987 NSF report’s1 key em- are now feasible in real time, benefiting tors are relatively heavy and expen- phases was the need for a national in- visualization users who need to explore sive. However, recent breakthroughs frastructure to enable visualization large data sets. in flat-panel and projector technology research and application. Many of the Developers have also made great have broken the CRT’s stranglehold. report’s needs have been satisfied, while advances in visualization-specific hard- The combination of low cost, high others remain a challenge. ware. is extremely resolution, and freedom from CRTs’

68 COMPUTING IN SCIENCE & ENGINEERING VIRUS STRUCTURE

Determining a virus’s 3D structure is the first step toward understanding its virulent func- tion. A new experimental approach uses cryo-electron microscopy (cryo-EM) for elucidating single-particle macromolecular structures (such as viruses) at the highest (1) (2) (3) quasi-atomic resolution, typically around 10 Å. Figures C1 through C3 show visualizations using combined surface and volume render- ing of the same half-shell model of the Rice Dwarf Virus (RDV).1 The different colors show the nucleo-capsid shell from the outside (Fig- ures C1 and C3) and the inside (Figure C2), elucidating the local and global complexity of (4) (5) (6) each structural protein unit’s quasi-symmetric packing. You can see each unit, exhibiting a trimeric fold, further visualized from two dif- Figure C. The Rice Dwarf Virus. These visualizations combine surface ferent views (Figures C4 and C5), as well as and volume rendering for the same half-shell model of the virus. The with different colors (Figure C6), to show the different colors show the nucleo-capsid shell from (1 and 2) the three different conformations of the mono- outside and (3) the inside. Each structural protein unit is further meric protein chain forming the trimeric visualized from (4 and 5) two different views as well as with (6) structure unit. All of the structure units are au- different colors. tomatically segmented from a reconstructed 3D electron-density . Previously, struc- tural biologists have largely attempted these 3D ultrastruc- sent significant computational and experimental challenges ture elucidation steps manually. The quasi-atomic models of in this research. these viruses and other macromolecular machines, con- structed via this structure-elucidation pipeline, provide mi- Reference crobiologists and drug-discovery scientists with crucial 1. Z. Yu and C. Bajaj, “Automatic Ultra-Structure Segmentation of Re- insight into molecular interactions within macro-assemblies. constructed Cryo-EM of Icosahedral Viruses,” IEEE Trans. Image Such complexes’ large physical size and complexity, com- Processing: Special Issue on Molecular and Cellular Bioimaging, vol. 14, bined with cryo-EM’s very low signal-to-noise ratio, still pre- no. 9, 2005, pp. 1324–1337.

weight and bulk constraints have led hundreds of megapixels by tiling pro- networked, can be used as control to an explosion of computer-driven jectors’ output. Although active sur- panels for a shared large display. displays in new contexts that go be- faces will still be relatively expensive yond simply replacing a user’s CRT in the near term, a longer-term vision Networking with a sleek flat-panel display that has includes gigapixel displays that will The visualization field has benefited a smaller footprint. eventually be as cheap, lightweight, greatly from the Internet’s expansion The primary limitation in interac- and ubiquitous as wallpaper. Physi- and commoditization. However, as tive visualization interfaces is the cally large displays that encompass an data sets continue to grow, it’s still dif- number of available pixels; we’re observer’s entire viewing field enable ficult—and sometimes prohibitive—to pixel-bound, not CPU-bound or even applications that use peripheral vision move large-scale data sets across even render-bound, and currently face a as well as the foveal vision we cur- the fastest networks to visualize data shortage. High-resolution displays rently use with medium-sized desktop locally. As such, we need advances will let us investigate new and exciting displays. Small-gadget displays will both in networking and remote and areas of the interface design space as have the one-megapixel resolution collaborative visualization algorithms, displays approach paper-like resolu- that we typically associate with desk- such as view-dependent algorithms, tion. We can create wall-sized displays top displays. Small handhelds are image-based rendering, multiresolu- with a resolution of dozens or even ubiquitous and portable, and, when tion techniques, importance-based

JULY/AUGUST 2006 69 V ISUALIZATION C ORNER

THE VISUALIZATION TOOLKIT n 1993, three visualization re- I searchers from General Electric’s corporate R&D center began to de- velop an open-source visualization system, known as the Visualization Toolkit (VTK). VTK was initially envi- sioned as a teaching and research collaboration tool, hence its release under an open-source license. The software gained rapid acceptance, in part due to the sophistication of its object-oriented design and software process, but also because of the user community that formed around it. VTK is now used worldwide and has helped spawn several small compa- nies and derivative products. Kitware (www.kitware.com), for example, Figure D. The VolView visualization software toolset. This example shows VolView was formed in 1998 to support VTK, being used to visualize the spinal column, blood vessels, and tissue slides. (Figure subsequently creating products courtesy of Kitware.) based on the toolkit, including the open-source ParaView parallel visual- ization system and the proprietary VolView volume-rendering application.1 VTK continues to evolve with contributions from researchers in academia, US national laboratories, and businesses, and is used in dozens of commercial software applications.

methods, and adaptive resource-aware turing visualization field. With open- source business model, in which cus- algorithms. source models, the user community it- tomers can pay for support and cus- self sometimes takes over some or all tomization while the community Visualization Software of the support burden. Releasing soft- continues to develop the free codebase. Dataflow toolkits, introduced more ware doesn’t have to be a gargantuan Similarly, the Insight Segmentation and than 15 years ago, heralded the first task; often, people who find that a par- Registration Toolkit (ITK; www.itk.org) generation of general-purpose visual- ticular piece of research software was an NIH open-source software ini- ization software. Various application- closely matches their needs are happy tiative intended to support a worldwide specific and open-source packages have to use software that is less polished community in image processing and appeared since, and many are in wide- than a commercial product. data analysis. It can interface openly spread use today. Several are delineated The Visualization Toolkit (VTK; with visualization platforms such as in the new VRC report.2 www..org) system4 began as an open- VTK and SCIRun, the University of A trade-off exists between quickly source initiative within General Elec- Utah’s visualization system, which has creating a one-off prototype that suf- tric’s Global Research division, and has also moved to open-source infrastruc- fices for a research paper but is too rapidly moved into mainstream use in ture software to ease its integration into brittle for general use and devoting the universities, national laboratories, and public and private research. time to create releasable code at the ex- industrial research labs worldwide (see pense of progressing on the next the “Visualization Toolkit” sidebar for Funding research project. One benefit for re- examples). It continues to accelerate de- We note with particular concern the searchers of releasing code to a user velopment by providing reusable soft- recent article by ACM President David community is that real-world use typi- ware, relieving programmers from Patterson5 that documents the decline cally spawns new research challenges reinventing necessary infrastructure. in both industrial and military funding strongly tied to real problems. Such The spin-off company Kitware (www. for basic research. Moreover, an in- ties are extremely important for a ma- kitware.com) is built around an open- creasingly low percentage of NSF pro-

70 COMPUTING IN SCIENCE & ENGINEERING Figure F. The Visualization Toolkit (VTK). This image FIgure E. The ParaView toolset. Here, ParaView is being shows the flow of liquid oxygen across a flat plate with a used to visually analyze a computational fluid dynamics cylindrical post perpendicular to the flow. (Figure data set. (Figure courtesy of Kitware.) courtesy of Kitware.)

Figure D demonstrates volume rendering using VolView, whereas Figure E shows a computational fluid dynamics visualization us- ing ParaView. Figure F uses the LOx Post data set, simulating the flow of liquid oxygen across a flat plate with a cylindrical post per- pendicular to the flow. This model studies the flow in a rocket engine, where the post promotes the mixing of the liquid oxygen.2

References

1. W.J. Schroeder, K. Martin, and B. Lorensen, The Visualization Toolkit: An Object Oriented Approach to Computer Graphics, 3rd ed., Kitware, 2004. 2. S.E. Rogers, D. Kwak, and U.K. Kaul, “A Numerical Study of Three-Dimensional Incompressible Flow around Multiple Posts,” Proc. AIAA Aerospace Sciences Conf., AIAA paper 86-0353, 1986, pp. 1–12.

posals are getting funding, and there separate workshops held in Bethesda, ducibility in technical and scientific appears to be a bias in proposal review- Maryland, in September 2004 and Salt developments. Such agency leadership ing toward low-risk incremental work Lake City, Utah, in May 2005. Based is critical if we’re to meet US needs in rather than attacking grand challenges. on our panelists’ deliberations, we critical areas and maintain competi- One important exception to this make the following recommendations. tiveness worldwide. downturn is the new US National Vi- sualization and Analytics Center Principal Agency Short-Term Policy (NVAC) initiative,6 which will spend Leadership Recommendation Recommendation several hundred million dollars in the NSF and NIH must make coordinated NIH and NSF can immediately imple- next five years on . investments in visualization to address ment funding policy changes to en- Much of this funding will be focused the 21st century’s most important courage evaluation of visualization and on the national security domain. Al- problems, which are predominantly collaboration between visualization and though considerable fundamental re- collaborative and cross disciplines, other fields, without requiring new search will arise from this short-term agencies, and sectors. Both NSF and funding initiatives. Likewise, journals program, NIH and NSF must ensure NIH can and should provide leader- and conferences should update their that the health and science domains are ship to other federal funding partners publication review criteria to reflect sufficiently funded, and that long-term and to the research communities they how important evaluation is in de- research continues to put the field on a support. To achieve this, they should termining techniques’ success and solid scientific foundation. modify their programmatic practices characterizing their suitability. Meth- to better engage visualization capabil- odological rigor should be expected Recommendations ities across disciplines important to when user studies are proposed or re- We explored these infrastructure and scientific and social progress, and en- viewed, and as necessary, visualization funding findings, as well as other find- courage and reward interdisciplinary researchers should collaborate with ings, with help from area experts in two research, open practices, and repro- those trained in fields such as human–

JULY/AUGUST 2006 71 V ISUALIZATION C ORNER

computer interaction, psychology, and other socially important concerns. This Jerry Prince, Will Schroeder, Jack statistics. Peer review of proposals and investment is critical if the US wants to Snoeyink, John Stasko, Barbara Tver- publications should also reward visual- remain competitive in a global research sky, Matthew O. Ward, Colin Ware, ization driven by real-world data and and development community that has Ross Whitaker, Turner Whitted, and tasks, in close collaboration with tar- increasing resources. In addition to Jarke van Wijk. get users. In other fields, review funding transitional research, such protocols should encourage domain programs should emphasize founda- References scientists who create partnerships with tional research and integration of 1. B.H. McCormick, T.A. DeFanti, and M.D. visualization researchers, just as en- methodologies from other fields, and Brown, Visualization in Scientific Computing, Nat’l Science Foundation, 1987. gagement with statisticians is consid- collaboration with domain specialists 2. C. Johnson et al., NIH-NSF Visualization Re- ered normal practice in areas such as who provide driving problems in areas search Challenges Report, IEEE Press, 2006. biomedical research. of national concern. A long-term 3. P. Lyman and H.R. Varian, How Much Infor- funding commitment is required to mation, 2003; www.sims.berkeley.edu/how Midterm Direction create and maintain curated data col- -much-info-2003. Recommendation lections and open-source software to 4. W. Schroeder, K. Martin, and B. Lorensen, The Visualization Toolkit: An Object-Oriented NIH and NSF should establish pilot promote open science. Characterizing Approach to Computer Graphics, 3rd ed., Kit- programs to combine efforts and cre- how and why visualizations work, sys- ware, 2003. ate collaborative development between tematically exploring visual represen- 5. D.A. Patterson. “The State of Funding for New Computer Science Initiatives in Com- puter Science and Engineering,” Comm. We recommend investment in a spectrum ACM, vol. 48, no. 4, 2005, p. 21. 6. J.J. Thomas and K.A. Cook, eds., Illuminating of centralized and distributed research programs. the Path: The Research and Development Agenda for Visual Analytics, Nat’l Visualization and Analytics Center, 2005. visualization and other research do- tations’ design space, developing new mains. Funding for such programs interaction approaches, and exploiting Robert Moorhead is the Billie J. Ball Profes- should contribute proportionately to the possibilities of novel display hard- sor of Electrical and Computer Engineering both the visualization research and the ware will be particularly important at Mississippi State University (MSU) and the domain specialty. This should help im- areas of emphasis. director of the Visualization, Analysis, and prove the penetration of emerging Imaging Lab. He is also the associate direc- technologies into new domains, in- Acknowledgments tor for research for the MSU GeoResources creasing their facility to move data and We’re indebted to the expert and vi- Institute. His research interests are focused share results through visualization. sionary VRC workshop panelists for around visualization of enormous data sets, Awards in this area should be dedi- sharing their considerable talents and remote visualization, and the effectiveness cated to open access of source code, insights, and to NIH and NSF for their of visualization techniques. Moorhead has a availability of research data to the sponsorship. In particular, we thank PhD in electrical and computer engineering worldwide community, and repro- panelists Maneesh Agrawala, Liz Bul- from North Carolina State University. Con- ducibility of the technical and scien- litt, Steve Cutchin, David Ebert, tact him at [email protected]. tific developments. Thomas Ertl, Steve Feiner, Felice Frankel, Bob Galloway, Alyssa Good- Chris Johnson directs the Scientific Com- man, Mike Halle, , puting and Imaging (SCI) Institute at the or the long term, we recommend a Chuck Hansen, Helwig Hauser, Karl University of Utah where he is a Distin- F coordinated and sustained national Heinz Höhne, Ken Joy, Arie Kaufman, guished Professor of Computer Science and investment in a spectrum of centralized Daniel Keim, David Laidlaw, Ming holds faculty appointments in the Depart- and distributed research programs to Lin, Leslie Loew, Bill Lorensen, ments of Physics and Bioengineering. His re- promote foundational, transitional, and Wayne Loschen, Patrick Lynch, Kwan search interests are in the areas of scientific applied visualization research in sup- Liu Ma, Alan MacEachren, Chris computing and scientific visualization. John- port of science, medicine, business, and North, Art Olson, Catherine Plaisant, son has a PhD in computer science from the

72 COMPUTING IN SCIENCE & ENGINEERING University of Utah. He founded the SCI Re- MA. He is the chief architect of VolumePro, and interactive representations and inter- search Group in 1992, which has since Mitsubishi Electric’s real-time volume ren- faces, and the experimental validation of vi- grown to become the SCI Institute, employ- dering hardware for PCs. His research inter- sualization techniques. Rheingans has a PhD ing more than 100 faculty, staff and stu- ests include computer graphics, scientific in computer science from the University of dents. Contact him at [email protected]. visualization, and graphics architectures. North Carolina, Chapel Hill. Contact her at Pfister has a PhD in Computer Science from [email protected]. Tamara Munzner is an assistant professor the State University of New York, Stony of computer science at the University of Brook. His work spans a range of topics, in- Terry S. Yoo is head of the Program for 3D British Columbia. Her current research in- cluding point-based graphics, appearance Informatics in the Office of High Perfor- terests are information visualization, graph modeling and acquisition, computational mance Computing and Communications, drawing, and dimensionality reduction. photography, 3D television, and face mod- National Library of Medicine, NIH. His re- Munzner has a PhD in computer science eling. Contact him at pfi[email protected]. search explores the processing and visual- from Stanford University. She was the IEEE ization of 3D medical data, interactive 3D Symposium on Information Visualization (In- Penny Rheingans is an associate professor graphics, and computational geometry. Yoo foVis) Program/Papers cochair in 2003 and of computer science at the University of has a PhD in computer science from the Uni- 2004. Contact her at [email protected]. Maryland Baltimore County. Her current re- versity of North Carolina, Chapel Hill. He is search interests include uncertainty in visu- the project officer who conceived and man- Hanspeter Pfister is associate director and alization, volume visualization, information aged the development of ITK, the Insight senior research scientist at Mitsubishi Elec- visualization, perceptual and is- Toolkit, under the Visible Human Project. tric Research Laboratories in Cambridge, sues in graphics and visualization, dynamic Contact him at [email protected].

ADVERTISER / PRODUCT INDEX MAY/JUNE 2006

Advertiser Page Number Advertising Personnel Marion Delaney Sandy Brown Krell Institute 7 IEEE Media, Advertising Director IEEE Computer Society, Phone: +1 212 419 7766 Business Development Manager Fax: +1 212 419 7589 Phone: +1 714 821 8380 Prentice Hall 44–45 Email: [email protected] Fax: +1 714 821 4010 Marian Anderson Email: [email protected] Advertising Coordinator Boldface denotes advertisements in this issue. Phone: +1 714 821 8380 Fax: +1 714 821 4010 Email: [email protected] Advertising Sales Representatives Mid Atlantic (product/recruitment) Midwest (product) Midwest/Southwest (recruitment) Northwest/Southern CA (recruitment) Dawn Becker Dave Jones Darcy Giovingo Tim Matteson Phone: +1 732 772 0160 Phone: +1 708 442 5633 Phone: +1 847 498-4520 Phone: +1 310 836 4064 Fax: +1 732 772 0164 Fax: +1 708 442 7620 Fax: +1 847 498-5911 Fax: +1 310 836 4067 Email: [email protected] Email: [email protected] Email: [email protected] Email: [email protected] Will Hamilton New England (product) Phone: +1 269 381 2156 Southwest (product) Japan Jody Estabrook Fax: +1 269 381 2556 Josh Mayer Tim Matteson Phone: +1 978 244 0192 Email: [email protected] Phone: +1 972 423 5507 Phone: +1 310 836 4064 Fax: +1 978 244 0103 Joe DiNardo Fax: +1 972 423 6858 Fax: +1 310 836 4067 Email: [email protected] Phone: +1 440 248 2456 Email: [email protected] Email: [email protected] Fax: +1 440 248 2594 New England (recruitment) Email: [email protected] Northwest (product) Europe (product) John Restchack Peter D. Scott Hilary Turnbull Phone: +1 212 419 7578 Southeast (recruitment) Phone: +1 415 421-7950 Phone: +44 1875 825700 Fax: +1 212 419 7589 Thomas M. Flynn Fax: +1 415 398-4156 Fax: +44 1875 825701 Email: [email protected] Phone: +1 770 645 2944 Email: [email protected] Email: [email protected] Fax: +1 770 993 4423 Connecticut (product) Email: fl[email protected] Southern CA (product) Stan Greenfield Marshall Rubin Phone: +1 203 938 2418 Southeast (product) Phone: +1 818 888 2407 Fax: +1 203 938 3211 Bill Holland Fax: +1 818 888 4907 Email: [email protected] Phone: +1 770 435 6549 Email: [email protected] Fax: +1 770 435 0243 Email: [email protected]

JULY/AUGUST 2006 73 The author has requested enhancement of the downloaded file. All in-text references underlined in blue are linked to publications on ResearchGate.